AO^QO-'Sk ''cyj COMPUTER SCIENCE & TECHNOLOGY ^ ACCESSING INDIVIDUAL RECORDS FROM PERSONAL DATA FILES USING NON-UNIQUE IDENTIFIERS ^ 0F % V £ \ \ °^A\i o? y NBS Special Publication 500-2 U.S. DEPARTMENT OF COMMERCE National Bureau of Standards NATIONAL BUREAU OF STANDARDS The National Bureau of Standards 1 was established by an act of Congress March 3, 1901. The Bureau's overall goal is to strengthen and advance the Nation's science and technology and facilitate their effective application for public benefit. To this end, the Bureau conducts research and provides: (1) a basis for the Nation's physical measurement system, (2) scientific and technological services for industry and government, (3) a technical basis for equity in trade, and (4) technical services to promote public safety. The Bureau consists of the Institute for Basic Standards, the Institute for Materials Research, the Institute for Applied Technology, the Institute for Computer Sciences and Technology, and the Office for Information Programs. THE INSTITUTE FOR BASIC STANDARDS provides the central basis within the United States of a complete and consistent system of physical measurement; coordinates that system with measurement systems of other nations; and furnishes essential services leading to accurate and uniform physical measurements throughout the Nation's scientific community, industry, and commerce. The Institute consists of the Office of Measurement Services, the Office of Radiation Measurement and the following Center and divisions: Applied Mathematics — Electricity — Mechanics — Heat — Optical Physics — Center for Radiation Research: Nuclear Sciences; Applied Radiation — Laboratory Astrophysics - — Cryogenics - — Electromagnetics 2 — Time and Frequency 2 . THE INSTITUTE FOR MATERIALS RESEARCH conducts materials research leading to improved methods of measurement, standards, and data on the properties of well-characterized materials needed by industry, commerce, educational institutions, and Government; provides advisory and research services to other Government agencies; and develops, produces, and distributes standard reference materials. The Institute consists of the Office of Standard Reference Materials, the Office of Air and Water Measurement, and the following divisions: Analytical Chemistry — Polymers — Metallurgy — Inorganic Materials — Reactor Radiation — Physical Chemistry. THE INSTITUTE FOR APPLIED TECHNOLOGY provides technical services to promote the use of available technology and to facilitate technological innovation in industry and Government; cooperates with public and private organizations leading to the development of technological standards (including mandatory safety standards), codes and methods of test; and provides technical advice and services to Government agencies upon request. The Insti- tute consists of the following divisions and Centers: Standards Application and Analysis — Electronic Technology — Center for Consumer Product Technology: Product Systems Analysis; Product Engineering — Center for Building Technology: Structures, Materials, and Life Safety; Building Environment; Technical Evalua- tion and Application — Center for Fire Research: Fire Science; Fire Safety Engineering. THE INSTITUTE FOR COMPUTER SCIENCES AND TECHNOLOGY conducts research and provides technical services designed to aid Government agencies in improving cost effec- tiveness in the conduct of their programs through the selection, acquisition, and effective utilization of automatic data processing equipment; and serves as the principal focus within the executive branch for the development of Federal standards for automatic data processing equipment, techniques, and computer languages. The Institute consists of the following divisions: Computer Services — Systems and Software — Computer Systems Engineering — Informa- tion Technology. THE OFFICE FOR INFORMATION PROGRAMS promotes optimum dissemination and accessibility of scientific information generated within NBS and other agencies of the Federal Government; promotes the development of the National Standard Reference Data System and a system of information analysis centers dealing with the broader aspects of the National Measurement System; provides appropriate services to ensure that the NBS staff has optimum accessibility to the scientific information of the world. The Office consists of the following organizational units: Office of Standard Reference Data — Office of Information Activities — Office of Technical Publications — Library — Office of International Relations — Office of International Standards. 1 Headquarters and Laboratories at Gaithersburg, Maryland, unless otherwise noted; mailing address Washington, D.C. 20234. - Located at Boulder, Colorado 80302. o. <3 & a. c3 COMPUTER SCIENCE & TECHNOLOGY: Accessing Individual Records From Personal Data Files Using Non-Unique Identifiers Gwendolyn B. Moore John L. Kuhns Jeffrey L. Trefftzs Christine A. Montgomery Operating Systems, Inc. 21031 Ventura Boulevard Woodland Hills, California 91364 Prepared for the Institute for Computer Sciences and Technology National Bureau of Standards Washington, D.C. 20234 ^ W ^ ™ U.S. DEPARTMENT OF COMMERCE Dr. Betsy Ancker-Johnson, Assistant Secretary for Science and Technology NATIONAL BUREAU OF STANDARDS, Ernest Ambler, Acting Director Issued February 1977 Reports on Computer Science and Technology The National Bureau of Standards has a special responsibility within the Federal Government for computer science and technology activities. The programs of the NBS Institute for Computer Sciences and Technology are designed to provide ADP standards, guidelines, and technical advisory services to improve the effectiveness of computer utilization in the Federal sector, and to perform appropriate research and development efforts as foundation for such activities and programs. This publication series will report these NBS efforts to the Federal computer community as well as to interested specialists in the academic and private sectors. Those wishing to receive notices of publications in this series should complete and return the form at the end of this publication. National Bureau of Standards Special Publication 500-2 Nat. Bur. Stand. (U.S.), Spec. Publ. 500-2,203 pages (Feb. 1977) CODEN: XNBSAV Library of Congress Cataloging in Publication Data Main entry under title: Accessing individual records from personal data files using non- unique identifiers. (Computer science & technology) (NBS special publication ; 500-2) Supt. of Docs, no.: CI 3. 10:500-2 1. Personnel records— Data processing. 2. Civil service— United States— Personnel management. I. Moore, Gwendolyn B. II. Institute for Computer Sciences and Technology. III. Series. IV. Series: United States. National Bureau of Standards. Special publication ; 500-2. QC100.U57 no. 500-2 [JK.766.5] 602Ms [353.001] 76-57950 U.S. GOVERNMENT PRINTING OFFICE WASHINGTON: 1977 For sale by the Superintendent of Documents, U.S. Government Printing Office, Washington, D.C 20402 (Order by SD Catalog No. C13:10:500-2). Stock No. 003-003-01726-4 Price $2.65 (Add 25 percent additional for other than U.S. mailing). FOREWORD How does a Federal agency locate a personal record in one of its massive files with only a limited amount of information for guidance? The Privacy Act of 1974 presented numerous agencies with just this problem when it specified: "It shall be unlawful for any Federal, state, or local government agency to deny to any individual any right, benefit, or privilege provided by law because of such individual's refusal to disclose his Social Security account number." Those Federal agencies denied the use of Social Security numbers are not the only ones interested in the solution to this problem. For instance, one might be required to locate individuals in a file who have been inadvertently or illegally recorded under more than one account number. Similarly, the results of a study of methodologies for retrieving an individual's record without the use of a universal identifier can be used to determine the threats to an individual's privacy. This report was written in response to this growing need for retrieving information using non-unique identifiers. Presented are selected methodologies for assisting Federal agencies in selecting retrieval algorithms and name lookup techniques; in analyzing their data by the identification of weighting factors and statistical sampling for determining error and omission rates; and lastly, predicting the accuracy and efficiency of candidate retrieval keys. Seymour Jeffery, Chief Systems and Software Division Institute for Computer Sciences and Technology National Bureau of Standards 111 ACKNOWLEDGMENTS The authors would like to express their appreciation to all those who contributed to the preparation of this document through their willingness to discuss their relevant experience and expertise. Particular thanks are due two members of the NBS Institute for Computer Sciences and Technology: Dr. Thomas C. Lowe, who coordin- ated this effort from its inception, gave incentive and direction to the project while assisting in eyery way possible with preparation of the final report; and John L. Berg, whose comments and suggestions have been invaluable throughout the course of the project. IV TABLE OF CONTENTS Page 1.0 Introduction 1 1.1 The Legislation 1 1.2 Retrieval of Individual Records 3 1.3 Statistical vs. Dossier Files 4 1.4 Privacy vs. Security 6 1.5 Summary 8 2.0 State-of-the-Art Retrieval Techniques 9 2.1 Introduction . 9 2.2 Varying Selective Values of Non-Unique Identifiers ... 10 2.3 Importance of the Name Variable in Retrieval Through Use of Non-Unique Identifiers 12 2.4 Selected State-of-the-Art for Name Lookup Techniques . . 12 2.5 Implementation of Name Lookup Techniques 26 2.6 Other Search Strategies 29 3.0 Identification of Weighting Factors 32 3.1 Accuracy Constraints Imposed by the Privacy Act 32 3.2 Determination of Weighting Factors for Existing Files. . 35 3.2.1 Data Base Definition and Forms Design 36 3.2.2 Data Collection Techniques 37 3.2.3 Keying of Data and Input Controls 42 3.2.4 Computer Edit Routines 44 3.2.5 Manual Checks for Accuracy 47 3.3 Summary of Weighting Factors 49 4.0 File Validation 52 4.1 Introduction 52 4.2 Validation Techniques 52 4.3 File Validation Example 53 5.0 Computer Simulation of Data Base Retrieval 59 5.1 Precision and Recall 59 5.2 Query Formulation for Best Results 61 5.3 Soft-Match Techniques 62 5.4 Example of Use of File Validation Summary Data ..... 62 5.5 Precision and Probable Match ........ 66 6.0 Conclusions and Recommendations 69 6.1 Conclusions 69 6.2 Recommendations 70 6.2.1 Evaluation of Current File Access Capability. . . 71 6.2.2 Evaluation of the Data Base 71 Appendix I - Precision Tables 73 TABLE OF ILLUSTRATIONS Page Table 2.4-1 IBM Alpha Inquiry System Personal Name Encoding Algorithm 15 Table 2.4-2 Western Airlines Surname Match Rating Algorithm. ... 18 Table 2.4-3 Rules for Developing Phonetic Frequency Codes 19 Table 2.4-4 Letter Distributions for Phonetic Frequency Code Name Lookup Technique 21 Table 2.4-5 Illustration of the Application of the NYSIIS Coding Techniques 22 Table 2.4-6 NYSSIS Name Coding Rules 23 Form 3.1-1 Decay Factors of Standard Biographical Data Elements 34 Form 3.2.1-1 Inherent Characteristics of Biographical Data Elements 38 Form 3.2.1-2 Data Base Definition Weighting Factors for Biographical Data Elements ..... 39 Form 3.2.2-1 Data Collection Procedure Weighting Factors for Biographical Data Elements 43 Form 3.2.3-1 Keying and Input Control Weighting Factors for Batch Input . 45 Form 3.2.3-2 Keying and Input Control Weighting Factors for Interactive Input 46 Form 3.2.4-1 Computer Editing Weighting Factors 48 Form 3.2.5-1 Weighting Factors for Manual Accuracy Checks 50 Form 3.3-1 Cumulative Weighting Factors for Standard Biographical Data Elements 51 Table 4.2-1 Suggested Format for File Validation Record 54 Table 4.3-1 Source Documents for Ten Sample Records from the Dummy Data Base. ............ 55 Table 4.3-2 Computer Listings for Ten Sample Records from the Dummy Data Base 56 Table 4.3-3 File Validation Record for a Sample Record from the Dummy Data Base 57 Table 4.3-4 Summary of Errors and Omissions Detected in Analysis of 10% of the File. 58 Table 5.3-1 Retrieval Ratios (Estimates of Proportion of File Retrieved) 63 Table 5.3-2 Statistical Classification of Surname Records as Compiled from Social Security Administration Accounts. 64 vi TABLE OF CONTENTS (continued) Page Appendix II - A Probabilistic Formulation of the Non-Unique Access Problem 174 Appendix III - Soft-Match Techniques 181 Appendix IV - FORTRAN Source Code for the Probability Mode. ... 186 VI 1 Digitized by the Internet Archive in 2013 http://archive.org/details/accessingindividOOmoor ACCESSING INDIVIDUAL RECORDS FROM PERSONAL DATA FILES USING NON-UNIQUE IDENTIFIERS G. B. Moore, J. L. Kuhns, J. L. Trefftzs, C. A. Montgomery Operating Systems, Inc. ABSTRACT The Privacy Act of 1974 places restrictions on the Federal, state, and local agencies' use of the Social Security account number as an identi- fier. For some agencies, compliance will involve changes in implementa- tion of retrieval algorithms. This report describes methodology applic- able to these changes in the more general context of the problem of retrieving individual records from files using non-unique identifiers. State-of-the-art retrieval techniques are discussed, a method for assign- ing reliability weights to various personal data elements is presented, file validation techniques for the error and omission rates of data items are suggested, and a retrieval probability model -- designed to show likelihood of retrieval of a subject's record given a variety of popula- tions, combinations of identifiers, and error/omission rates -- is des- cribed. A methodology is developed for forming confidence factors from the established error/omission rates for combinations of non-unique identifiers that are candidates for use as retrieval keys. Use of these confidence factors as indices into the precision tables produced by the probability model is described. Key Words: Data retrieval; file validation; name lookup; non-unique identifiers; personal data files; Privacy Act, probability model ; retrieval . 1.0 INTRODUCTION 1. 1 The Legislation The Privacy Act of 1974 opens with the words: "The Congress finds that- (1) the privacy of an individual is directly affected by the collection, maintenance, use, and dissemination of personal information by Federal agencies; (2) the increasing use of computers and sophisticated informa- tion technology, while essential to the efficient operations of the Government, has greatly magnified the harm to indi- vidual privacy that can occur from any collection, main- tenance, use, or dissemination of personal information; (3) the opportunities for an individual to secure employment, insurance, and credit, for his right to due process, and other legal protections are endangered by the misuse of certain information systems; 1 (4) the right to privacy is a personal and fundamental right protected by the Constitution of the United States; and (5) in order to protect the privacy of individuals identified "in information systems maintained by Federal agencies, it is necessary and proper for the Congress to regulate the collection, maintenance, use, and dissemination of infor- mation by such agencies."* There has been increasing public concern over the trend toward integra- tion and centralization of data banks containing personal information. In the privacy legislation, Congress has explicitly defined the rights of citizens whose records are maintained by Federal agencies as well as the responsibilities of the agencies handling the data. In addition, the legislation also established the Privacy Protection Study Commission which is specifically authorized to: "...make a study of the data banks, automated data processing programs, and information systems of governmental, regional, and private organizations, in order to determine the standards and procedures in force for the protection of personal infor- mation... In the course of conducting the study... the Commis- sion may research, examine, and analyze. . .the use of social security numbers, license plate numbers, universal identifiers, and other symbols to identify individuals in data banks and to gain access to, integrate, or centralize information systems and files." 2 Specific prohibitions regarding use of universal identifiers will, for the most part, await the results of the Privacy Protection Study Commission study. The current emphasis is rather on restricting the data maintained to that directly relevant to the agency. However, si-nce much of the concern over use of universal identifiers has been generated by the use of the Social Security account numbers and their potential for linking files, the Act has specifically provided against its use as a required item of information unless a Federal statute specifying such a disclosure is applicable. "It shall be unlawful for any Federal, State or local govern- ment agency to deny to any individual any right, benefit, or privilege provided by law because of such individual's refus- al to disclose his social security account number." 3 P rivacy Act of 1974 , Public Law no. 93-579, Section 1 2 Loc. cit. 3 Ibid. , Section 7. Thus the effect of the legislation is that individuals whose records are maintained in the files of Federal agencies must be allowed to review their own records on request, the retrieval must be possible without use of the Social Security account number, the data maintained in the files must be timely, and only data necessary in the performance of the agency's mission should be stored. For agencies that have relied heavily on the Social Security number for purposes of retrieval, this requirement will necessitate some major re- adjustments. Further restrictions are imposed on the type of information that can be solicited from individuals who have requested copies of their record by the July 1, 1975 guidelines from the Office of Management and Budget, "Guidelines for Implementing Section 3 of the Privacy Act of 1974," which states that the items published in the Federal Register by each agency must include: "...the information necessary to identify the record. Where the system employs a specialized identification scheme, the individual should not be required to provide such a number or symbol as an absolute requirement, although the individual might be requested to supply it if he or she can reasonably be expected to know it. Instead, alternative combinations of personal characteristics may be used to identify individ- uals who may have lost, forgotten, or are unaware of their identification numbers or symbols. For example, the combina- tion of name, date of birth, or place of birth, and father's first name may be sufficient to identify an individual with- out the use of a system identification number. "4 Thus, although use of record identification numbers, driver's license numbers, etc., is not prohibited, careful attention must be given to the selection of a set of identifiers that will be considered reasonable in the particular environment. 1.2 Retrieval of Individual Records The implications of these restrictions will differ from agency to agency. It is important, however, that those charged with responsibility for the selection and implementation of identification schemes be aware of the retrieval effectiveness of non-unique personal identifiers. The potential for retrieval by means of exclusive use of combinations of these keys should be understood, and the relevance of this potential for use with currently active data bases should be examined. Consideration must also be given to the question of the reliability of the data contained in these files since the success of the retrieval algorithm depends upon the accuracy of the data base itself. Errors may have been introduced into the data base (during data collection, data entry, data update, etc.), Office of Management and Budget Circular A-108 and accompanying "Guide- lines for Implementing Section 3 of the Privacy Act of 1974," Federal Register , 9 July 1975, Vol. 40, No. 32, pp. 28948-28978. 3 with the result that a request from an individual for a copy of his re- cord might fail due to the presence of erroneous data in the record itself. In this respect, retrieval algorithms based on Social Security account number or other standard universal identifiers have advantages (it is easier to alert clerical staff to the necessity of checking one or two numbers for accuracy than it is to train them to be careful with all biographical data), adequate retrieval is certainly possible using non-unique identifiers such as name, birthdate, etc. In fact, such retrieval is sometimes possible even when the variables most closely identified with an individual are removed from the records. Statisticians involved in the collection and analysis of statistical data for research purposes have been concerned with the problem of protection of the in- dividual whose data is maintained in the files. The need to collect such data in order to study causes of disease, poverty, violence, migra- tion and other sociological phenomena is well understood, but the individ- uals involved have the right to expect that information given in con- fidence will be treated with integrity. Dr. Fellegi of the Dominion Bureau of Statistics in Canada has summarized the problem: "The public benefits indirectly from the legitimate uses of statistics by governments, businesses, non-profit organiza- tions, academic users, etc.; yet, the public is also concerned about the increasing burden of providing the required statis- tics and about the real or imagined possibility of the misuse of the data provided by them. The explosive increase in the demands for more statistics can only be met, without impossible response burdens being put on the public, through a more effect- ive exploitation of the data. This increases geometrically the 5 magnitude of the problem of checking tabulations for disclosure." The seemingly contradictory requirements of protecting individual privacy while at the same time providing the citizen the benefits attendant on the introduction of automated record keeping systems create a burden for those responsible for maintaining the systems. They must restrict the data retained in the file to that clearly relevant to the application, but at the same time be able to identify correctly all individuals concerned on demand. Data must be available for research efforts, but information maintained to support these efforts must be restricted and accessing of individual records carefully monitored. The dilemma thus created is discussed further in the next section. 1.3 Statistical vs. Dossier Files Until recently, it was generally believed that a basic distinction ex- isted between statistical files (those with names, addresses, etc., removed from the individual record) and dossier files (those containing explicitly descriptive information about the individuals). Dossiers 5 Fellegi, I. P., "On the Question of Statistical Confidentiality," Journal of the American Statistical Association , 67 (March 1972), p. 7-18. 4 were considered dangerous from the point of view of exposure whereas statistical files were not. James Martin, in discussing this distinc- tion, says: "It is important to separate the concept of statistical files from the concept of dossiers about individuals. From the point of view of privacy protection, they are entirely differ- ent. Many of the uses to which the information in government and research data banks will be put are statistical. In statistical files the identification of the individual should be stripped off. Once that is done the files can be used more freely. It is still necessary to make sure that people cannot be identified by some other means, such as home location, a high salary, the fact that he has 13 children or cats, weighs 300 pounds, or has a combination of factors that makes him unique. It is sometimes possible to extract information about individuals from a statistical data bank by asking combinations of questions. Sometimes a large number of questions would be needed and controls can be devised to prevent such interrogation. The controls are more complex and subtle than on the type of data bank in which details about an individual are requested directly. In general, statistical data banks will not reveal the identities of the persons about whom data are recorded. "^ Recent research indicates that the data contained in statistical files is not necessarily protected from unauthorized disclosure. Dr. SchloYer of the Department of Medical Statistics at the University of Ulm in Germany gives the following account of his investigation of retrieval from inter- active data banks. "Until fairly recently, removing name, address, and -- per- haps -- date of birth from a record would secure anonymity in most cases. The advent of the computer has changed this situation. It was soon realized that the distinction between a dossier data bank (returning full identifying information to the user) and a statistical data bank (providing only numbers of persons) is largely delusive, if the user is allow- ed to communicate with the data bank by dialogue. A prerequisite for this intrusion by dialogue is preknow- ledge. The intruder must have some a_ priori information from the statistical data bank. The same information must be stored in Mr. X's record, so that preknowledge and the corresponding part of the record can be matched. The pre- knowledge must suffice to identify or, at least, nearly identify Mr. X. An identification experiment with authentic 6 Martin, James, Security, Accuracy, and Privacy in Computer Systems Prentice-Hall, New Jersey, 1973, p. 432. data will be described in this paper. It will be seen that relatively little preknowledge will do for identification purposes -- at least for the life 'attacked' in this experi- ment.'^ Not only does Dr. Schltirer's work make it clear that records can be re- trieved through the use of non-unique identifiers, it also gives added credence to the requirement that only relevant data be maintained in the files, since inclusion of extraneous data makes the record more vulnerable. Further indication of the need for such restrictions comes from the re- search in record linkage carried out by Drs. Fellegi and Sunter. They have developed "...a mathematical model... to provide a theoretical framework for computer-oriented solution to the problem of recognizing those records in two files which represent identical persons, objects or events (said to be matched) ...A comparison is to be made between the recorded characteristics and values in two records (one from each file) and a decision made as to whether or not the members of the comparison-pair represent the same person or event, or whether there is insufficient evidence to justify either of these decisions at stipulated levels of error... A theorem describing the construction and properties of the optimal linkage rule and two corollaries to the theorem which make it a practical working tool are given. "8 The existence of such techniques for linking individual records on the basis of non-unique identifiers again points up the feasibility of imple- menting retrieval algorithms based on those variables. An awareness of the availability of such techniques further clarifies the nature of the threats that currently exist to individual privacy. It also emphasizes the importance of the provisions for relevancy of data in the Privacy Act since obviously the more data available, the easier such linkages become. Similarly, the necessity for building audit checks into the system to provide a means for determining what information was accessed by whom becomes clear in the light of currently available methodologies. 1.4 Privacy vs. Security Privacy has been perceptively defined by Westin as the right "...to g determine what information about ourselves we will share with others,"' 7 Schlbrer, J., "Identification and Retrieval of Personal Records from a Statistical Data Bank," Methods of Information In Medicine , 14 (1, 1975). o Fellegi, I. P., and A. B. Sunter, "A Theory for Record Linkage," Journal of the American Statistical Association , (Dec. 1969), pp. 1183-1210. 9 Westin, Alan, as quoted by I. P. Fellegi, "On the Question of Statistical Confidentiality," op.cit. , p. 18. 6 where the 'others' are persons or agencies who are presumably authorized to hold and use such information. Thus the legislative concept of pri- vacy in effect concerns the protection of information about individuals from misuse by duly constituted authorities, or persons or agencies to which such authority has been delegated (e.g., for statistical studies). Security, on the other hand, is concerned with protection of this infor- mation from unauthorized 'others.' Such unauthorized access may be de- liberate, or inadvertent: e.g., a person seeking information from his own record may be given the dossier of another individual by mistake. Protection of information in a computerized file from unauthorized access is partly a technical, system-internal, matter, and partly a procedural, system-external matter. System-internal protections include obvious items such as use of passwords and other identifying techniques to guard against deliberate access attempts by unauthorized persons. More fundamentally, such protection requires a high degree of precision in the retrieval al- gorithm to prevent inadvertent delivery of the wrong information record to a particular requestor (as well as erroneous updates and purges). Procedural protections involve basic items such as locks and security guards, as well as requests for proof of identity from persons seeking information about their own records, further verification based on ques- tions about the content of the record before it is delivered to the re- questing individuals, and so forth. In the case of a computerized infor- mation bank where requesting individuals are allowed access via a computer terminal, a combination of technical and procedural safeguards is required to insure protection of the component information records. In a recent article on computer privacy and computer security, Willis Ware states: "In the context of computer-based systems, the matter of access control is an essential part of a larger issue referred to as computer security which can be defined as: The protection of the equipment, facilities and data of an information system against deliberate or acci- dental damage, and against denial of use by legitimate users, together with the assurance that information will be delivered by the system only to individuals authorized to receive it."^ The phrase "legitimate users" of an information system implies in the privacy context that legitimate uses of personal information can be de- fined, and that "individuals authorized" to access such information will respect the right to privacy of those persons who provided it. This is Ware, Willis H. , "Computer Privacy and Computer Security," Bulletin of the American Society for Information Science , Vol. 1, No. 3, Oct. 1974. 7 the core of the privacy issue, which essentially involves the loss of control over personal information incurred by an individual in providing such data to some agency or data bank, and the profound effect that re- linquishing control of this information can have on the life of the particular individual. It is the concern of those involved in the effort to protect the privacy of the individual to identify measures, legal, procedural, and technological, by which potentially detrimental effects of this loss of control can be nullified. 1.5 Summary In Section 1 some of the difficulties faced by those responsible for the management of personal data bases in Federal agencies are outlined. Pro- blems inherent in the provision of continued service with no attendant reduction in efficiency due to the increased emphasis on the protection of individual privacy have been stressed. The remainder of this report is concerned with the presentation of methodologies potentially useful in the minimization of these difficulties. The application of current retrieval technologies to the problem of retrieval through use of non-unique personal identifiers is the basic methodology under consideration. A discussion of relevant retrieval algorithms is given in Section 2. Section 3 suggests an approach to determination of the accuracy and reliability of the various identifiers contained in current files. Forms that might be used in carrying out an analysis procedure are introduced, and a mechanism for arriving at weighting scores for the individual identifiers is described. Section 4 demonstrates a similar mechanism for analyzing error and omission rates for each of the personal identifiers contained in the file. Section 5 describes the probability model developed to assist in deter- mination of the best combination of personal identifiers available for use as retrieval keys given the idiosyncratic error and omission rates of those variables within a particular file. An example of the use of the output from the probability model in making this determination is given. Conclusions are formulated in Section 6 which also contains a summary of procedures recommended for those involved in the analysis of personal data files. Four Appendices are included: Appendix I: Precision Tables Appendix II: A Probabilistic Formulation of the Non-Unique Access Problem Appendix III: Soft Match Techniques Appendix IV: FORTRAN Source Code for the Probability Model. 2.0 STATE-OF-THE-ART RETRIEVAL TECHNIQUES 2.1 INTRODUCTION The personal data files affected by the provisions of the Privacy Act represent a wide spectrum of applications with very different require- ments. By definition the files all contain dossier-type data, but they vary in terms of the non-biographical variables maintained, the ulti- mate use of the data, the frequency of file access, the retrieval mechanisms employed, the degree to which accuracy has been stressed, etc. These distinctions reflect the differing operational requirements of the agencies involved. Clearly, the provisions of the Privacy Act giving the individual access to his own records will cause re-evaluation of file management requirements and, in many cases, some system changes will be necessary in order to make this data available. A clear-cut requirement for some system changes exists when the retrieval keys do not include any of those combinations of identifiers that can be readily recalled by the individual. Similarly, systems that have placed a high degree of reliance on the use of the Social Security account number in the retrieval algorithms will require extension of these system modifications to insure effective retrieval. But even if these changes result in the availability of a standard set of biographical variables as potential retrieval keys, the level of accuracy maintained for those variables may fall short of that required of retrieval keys. It is entirely possible that a higher degree of accuracy exists for the vari- ables that have historically been used as retrieval keys than for those now promoted as candidates for that task. Such a disparity in levels of accuracy could have resulted from awareness on the part of the clerical staff of the need to verify the retrieval keys, or from the programming of special checks on those fields, or the like. Further, this skewing of levels of accuracy may well have gone unnoticed since the retrieval algorithm has operated sufficiently well to fulfill the operational requirements of the agency. To illustrate this point, consider the case where an agency uses a specific data item, for example AGE, to trigger notification of the individual concerning a change in status. In implementing and maintain- ing such a system, the emphasis on manual and system checks of the vari- ables BIRTHDATE and AGE would be greater than that on NAME and ADDRESS. Indeed, since postmen have become expert at detecting near matches on name and street numbers, the degree of accuracy required to complete the correspondence is indeed less than that required for retrieval of the record on the basis of current age. One of the authors, Mr. Trefftzs, has experienced this phenomenon. His is an unusual name that causes a great deal of confusion. As a result it is seldom spelled correctly. Mr. Trefftzs has, in recent weeks, received communications from a variety of sources (governmental, maga- zines, professional societies, etc.) in which his name was consistently 9 misspelled in a number of inconsistent ways: TREVITZ TRAFFT TRAFFTZS TRRFFTZA TREFTZ etc. These misspellings prevented neither the retrieval of his record nor the delivery of the message. Yet, if the systems involved were capable of handling retrieval on the basis of biographical data including name, and he were to request a copy of his record from each, it is likely that some of the systems would fail to locate the record, given the nature of the error that exists in the name itself. This emphasizes the importance of the change of system perspective attendant on intro- duction of new accessing requirements. Problems such as gross mis- spellings of names may block retrieval of certain records until the level of accuracy of the NAME variable has been brought within reason- able bounds. Existing file organizations and retrieval mechanisms may thus be entirely responsive to the needs of a particular agency while being inadequate in terms of handling the new requirements. The most difficult aspect of this problem arises from the fact that the required retrieval capabili- ties may not have been built into the original system. In some cases, extension of current system capabilities will be sufficient to handle the new requirements; in others, new sets of capabilities will be needed; still other situations will arise where restructuring of the files is required in order to make this data available in the required way. Decisions regarding the system changes required will be based on several factors: demand (i.e., the number of requests for this information), capabilities of the current system (some will be set up in such a way that required changes will be minimal), characteristics of the data base (number of non-unique identifiers, available accuracy rates, typical problems with a given variable), etc. If it is evident that the complete redesign of the system is required in order to meet the new regulations, the agency will want to examine data base structures and retrieval techniques that might assist in optimizing the new system. The re- mainder of this section is devoted to a description of the current state-of-the-art for retrieval techniques as they apply to the problem of accessing data using non-unique identifiers. Since there is no single answer to the implementation of these requirements, this section is presented in the form of a survey in the hope that the availability of this information will clarify the options available to those con- fronted with the Droblem. 2.2 VARYING SELECTIVE VALUES OF NON-UNIQUE IDENTIFIERS Some of the identifiers available in personal data files are highly individualized. The variable NAME is the most obvious example of this. Other data items for example, SEX, are binary, and therefore not particu- larly useful in identifying an individual record. Yet collection of even the most straightforward items can be difficult since the items requested are sometimes subject to vastly different interpretations. 10 There is a natural tendency to assume that basic biographical data is easily obtainable, and that standard definitions can be taken for grant- ed. However, in actual data collection situations, surprising misunder- standings can and do occur and, most importantly, are often not detected until much later. The result is that variables considered to be reliable in the file are many times, in reality, only marginally so. An interesting discussion of the problems inherent in obtaining biograph- ical data is given in the Uniform Hospital Discharge Data Demonstration Summary Report : "DATE OF BIRTH: Date of birth is usually an easy item to collect on admission. However, there are instances of elderly persons not remembering their exact date of birth, and others who provide incorrect dates for per- sonal reasons. SEX: The incidence of encountering a patient with true indeterminate sex status is wery low. Most hospitals feel the indeterminate category is unnecessary, but hospitals who perform trans-sexual operations find it a useful item. MARITAL STATUS: While this item is usually easy to ob- tain, there are some inherent difficulties in assuring validity, and these include: Persons living in common-law relationships who do not wish to be classified under any of the alterna- tives. Divorced persons who consider themselves single and state so at admission. Single females admitted for childbirth do not generally want to be identified as single and state they are married. The original intent of this item was to attempt to secure some indication of the living arrangement of a patient, but it proves to be of marginal utili- ty for this purpose. This is a complex variable to establish if the many socio-economic factors which enter into the underlying health care implica- tions associated with living arrangements are con- sidered. "H 11 Hodgson, D. A., L. E. Kucken, and J. M. Ensign, The Uniform Hospital Discharge Data Demonstration, Summary Report , Health Service Foundation Chicago, 1973, pp. 30-31. 11 2.3 IMPORTANCE OF THE NAME VARIABLE IN RETRIEVAL THROUGH USE OF NON-UNIQUE IDENTIFIERS Use, as well as inherent variability, of identifiers will vary greatly from one application to the next. The variability within groups of identifiers is probably most extreme for the NAME variable. The fol- lowing example of such variability comes from a bibliographic data base application. The report states that: "...many of the individual data elements exhibit great variety (i.e., lists of their contents are extensive), and show relatively disparate distributions. This be- havior is encountered in different degrees in regard to items such as words in the titles of monograph or periodical articles, assigned subject headings, authors' names and citations. ...In general, the distributions are approximately hyperbolic, so that a small portion of items may account for a substantial proportion of occurrences, while the majority of items occur only infrequently. -...Of all the data elements, personal author names exhibit a distribution which is at its most extreme in one direction. As is shown ...the most frequent (full name) author name in a file of 50,000 names occurred only sixteen times, while over 35,000 of the names, or over 70 percent of the file, occurred once only. "12 The same tables show that, with respect to surnames only 20,000 or 40% of the file, occurred only once. These figures emphasize the high selective value of NAME as an identifier and thus provide the rationale for the following consideration of name lookup techniques. 2.4 SELECTED STATE-OF-THE-ART NAME LOOKUP TECHNIQUES A majority of the name lookup algorithms currently in use are based on a phonetic scheme aimed at minimizing the common transcription problems in the recording of names. The IBM Alpha Search Inquiry Program des- cription states that: "Using the phonetic approach for personal name increases the user's ability to access the alpha search record even though the exact spelling of a name may not be known. 12 D. W. Fokker and M. F. Lynch, "Application of the Variety-Generator Approach to Searches of Personal Names in Bibliographic Data Bases- Part I. Microstructure of Personal Authors' Names," Journal of Library Automation , 7 (2 June 1974), pp. 105-117. 12 Thus the effects of transcription errors, partially illegible signatures on correspondence, and sound- alike names can be minimized."^ In analyzing name lookup techniques, it is useful to refer to the in- formation systems concepts of precision and recall. Recall represents the number of relevant items contained in a given file which are re- trieved in response to a particular query, while precision refers to the number of file items retrieved which are actually relevant to the query or search prescription. In general, the objective of system improvement is to increase the de- gree of precision or reliability, without substantially increasing the amount of irrelevant material recalled or selected from the file. In a document retrieval system, however, it is rare (in practice, at least) that there is a single file item which is capable of satisfying a given query. Generally speaking, several documents will be relevant to a query, where the degree of relevance is defined concretely in terms of the concepts used in the query and abstractly in terms of actual utility to the requestor. Conversely, in a name retrieval system, the usual case is that there is a unique file item which satisfies a query, although the existence of other items listed under alternate spellings of the name must also be assumed by the search strategy. This is a critical issue, which does not appear to be effectively handled by many of the name search tech- niques investigated in the course of this study. In evaluating name lookup techniques for which complete documentation is available, there are two essential criteria for determining relative effectiveness. Both of these may seem obvious, but since several of the lookup techniques described here fail on one or both counts, the topic merits discussion. The first basic question is simply: does the strategy allocate similar names to the same logical group -- and more particularly, are alternate versions of the same name (Heinz, Heintze) in the same logical group? The answer is often in the negative, and this is the primary means of effecting an improvement in reliability of a given method, as we shall consider below. The second basic evaluation criterion is the converse of the first: does the strategy allocate dissimilar names to different logical groups? Although the possibility of violating this principle appears unlikely, many name lookup techniques do in fact fail to satisfy this criterion, 13 Alpha Search Inquiry System General Information Manual , GH20-1188-3, IBM Corporation, White Plains, N.Y., April 1974, p. 7. 13 including Soundex, the first widely used phonological ly based system for name encoding. Most of the systems currently in use have been influenc- ed by the rules established 50 years ago by Margaret Odell and Robert Russell for the Soundex system. Knuth summarizes the original Soundex rules as follows: "1. Retain the first letter of the name, and drop all occurrences of a, e, h, i, o, u, w, y in other positions \ 2. Assign the following numbers to the remaining letters after the first: b, f , p, v ■* 1 1 + 4 c, g,j,k, q, s,x, z + 2 m, n + 5 d, t ■* 3 r •> 6 3. If two or more letters with the same code were adjacent in the original name (before Step 1), omit all but the first. 4. Convert to the form "letter, digit, digit, digit" by adding trailing zeros (if there are less than three digits), or by dropping rightmost digits (if there are more than three ).-"I4 This algorithm fails to meet the evaluation criteria described above, as Knuth has pointed out: "...the names Euler, Gauss, Hilbert, Knuth, Lloyd, and Lukasiewicz have the respective codes E460, G200, H416, K530, L300, L222. Of course this system will bring to- gether names that are somewhat different, as well as names that are similar; the same six codes would be ob- tained for Ellery, Ghosh, Heilbronn, Kant, Ladd, and Lissajous. And on the other hand a few related names like Rogers and Rodgers, or Sinclair and St. Clair, or Tchebysheff and Chebyshev, remain separate. But by and large the Soundex code greatly increases the chance of finding a name in one of its disguises. "15 The type of refinement that has taken place in development of name look- up techniques is obvious from comparing the results of applying the rules used in the IBM Alpha Search Inquiry System (Table 2.4-1) with those of Soundex, using two of the examples mentioned above. 14 Knuth, Donald E., The Art of Computer Programming, Vol. Ill: Sorting and Searching , Addison-Wesley, Reading, Mass., 1973, p. 392. 15 i5 Ibid. 14 Table 2.4-1 IBM Alpha Inquiry System Personal Name Encoding Algorithm 16 This routine builds a phonetic key from the last name passed by the calling program. The phonetic key is 7 packed bytes (14 digits). The routine uses two tables, a first character table and a basic table. The first character table (see below) is used to recognize first letters or letter combinations and give them significance. For example, A in Armour has a value of 1 to be put in the key. A in Farmer has no value and is ignored. If the first letter or letter combination does not appear in the first character table, position and the basic table is used the record key position. a zero is inserted in the first key to code the first character into The basic table (see below) is used to recognize letters or letter combinations that are phonetically equivalent. Vowels and the letters H, W, Y are ignored. Letters with the same phonetic value which are adjacent to each other are treated as single letters. For example, in CK and TT the second character is dropped. However: if a third charac- ter has the same value as the previous two, it is retained. The re- mainder of any unfilled key is filled with zeroes. The first character table used in this routine is: The basic table used in this routine is: Letter(s) A E GF GM GN H I J KN PF PN PS U W (except WR) WR Y Value 1 1 08 03 02 2 1 3 02 1 08 02 00 1 4 04 5 Code Letter(s) Z,S,CI,CY,CE,TS,TZ 1 D,T 2 N 3 M 4 R 5 L 6 J,SH,SCH,CH 7 C,G,K,Q,X,DG 8 F,V,PH 9 B,P Alpha Search Inquiry System General Information Manual , IBM Corpora- tion, White Plains, New York, GH20-1188-3, 1974, pp. 41-42. 15 Table 2.4-1 IBM Alpha Inquiry System Personal Name Encoding Algorithm (Continued) 16 EXCEPTIONS TO BASIC TABLE PROCESSING* In certain situations where letters or groups of letters have multiple sounds, a second and third pass through name encoding is made. Example: CH Has a hard sound in Nichols, a soft sound in Chavez. 1st Pass (Coded in Record) 2nd Pass 3rd Pass cz 70 6 CH 6 70 CK 7 7 6 C 7 7 6 K 7 7 6 DS 10 10 DZ 10 10 TS 10 10 TZ 10 10 Following are some examples of the results of coding personal names: NAME HARPER KEY 2 4 9 4 = 24940000000000 NAME COLLIER KEY 07 5 4 = 07540000000000 NAME S C H U L T Z KEY 06 5 = 06500000000000 NAME LIVINGSTON KEY 05 8 2701 2= 05827012000000 ^Occurrences of the letter combinations in this table cause generation of alternate phonetic keys. When this happens, the lookup algorithm will use both phonetic keys in the search for potential matches. This procedure alleviates many of the problems attributable to phonetic miss- spellings (e.g., Ohrbock for Ohrbach). A cross-reference table is used to connected sound-alike names (e.g., Lyle and Lisle) which are not handled by this algorithm. ; 16 SEARCH CRITERION SURNAME ASSIGNED VALUES SOUNDEX ALPHA Assign similar names to the same logical group RODGERS R326 04740000000000 ROGERS R262 04740000000000 Assign dissimilar names to differ- ent logical group KANT K530 02100000000000 KNUTH K530 07210000000000 In both these cases, the Alpha Search Inquiry System has solved the pro- blem. There are, however, an equally good set of examples for problems that still remain. Thus, although this algorithm represents progress over that used in the Soundex system, it is far from perfect. Sometimes quite straightforward algorithms are capable of handling this problem successfully for files of limited extent. The match rating approach used by Western Airlines (see summary in Table 2.4-2) would correctly identify RODGERS and ROGERS as a potential match. In the Western system, the matching algorithm would compare RDGERS with RGRS. The number of unmatched characters would be 1, the resulting similarity rating would be 5, and, hence, this record would be flagged for retrieval On the other hand, this algorithm would also consider KANT to be a ly match for KNUTH, which points up the limitations of the system. is used successfully in this application because only a small file searched at a time. Ms. Gail Hogan, Manager of Passenger Service Systems and Programming for Western, states: "Since our reservations system is flight/date oriented (i.e., we limit the scope of our search by date and/or flight number), we have little difficulty in obtaining an exact name match. However, if there are passengers on a flight with the same surname, we display a list of these to the agent. If an exact match is not made, we continue processing. . .in order to compile a similar surname list. "17 Another example of failure to meet the criterion of assigning dissimilar names to different buckets can be given using the Standardized Phonetic Frequency Code algorithm (see Table 2.4-3) as described in the NYSIIS evaluation of name search techniques. 18 In this system, an attempt was like- It is 17 18 Personal Communication from Gail Hogan, Manager Passenger Service Systems and Programming for Western Airlines, September 15, 1975. Taft, Robert L. , Name Search Technique , New York State Identification and Intelligence System, August 1970. 17 19 Table 2.4-2 Western Airlines Surname Match Rating Algorithm The input surname and each of the PNI [Personal Numeric Identifiers] - item surnames are encoded by this program prior to comparison. Encoding is as follows: (1) Deletion of all vowels unless the vowel is the first character of the surname. (2) Elimination of all double consonants by deleting the second contig- uous usage of any consonant. (3) Reducing all encoded names to a maximum of six characters. This is done by retaining the first three and last three encoded char- acters. The lengths of each pair of encoded names (input and PNI item) are examined.* If they differ in length by more than two, no similarity comparison is performed. A minimum acceptable similarity rating is established for each pair of encoded names as follows: Sum of lengths is 4 or less; rating of 5 Sum of lengths is 7 or less; rating of 4 Sum of lengths is 11 or less; rating of 3 Sum of lengths is 12; rating of 2 The minimum acceptable rating establishes which PNI surnames are not "similar enough" to be considered by the agent. Comparison of encoded names is then performed. This comparison is from left to right, character by character. Matching pairs of characters are deleted. This comparison continues until either encoded name has no re- maining characters. All unmatched characters in both encoded names are packed to the right and comparison proceeds from right to left. On completing these comparisons the number of unmatched characters in the longer name is subtracted from a value of six with the result being the similarity rating for that PNI item. Each PNI item that has a similarity rating equal to or greater than the minimum rating established for this time will be added to the Retrieval Control Record [RCRJ. This program will add to the RCR until the record is filled. At that point additional entries will replace other items in the RCR wnich have lower similarity ratings. If no item with a lower rating is found, the PNI item will be ignored. *That is, the length of the encoded input name is compared with that of the encoded name in the file. 19 Gail Hogan, op.cit. 18 Table 2.4-3 Rules for Developing Phonetic Frequency Codes NAME FIELD CODE J.KUHNS G.ALTSCHULER 1. In the name field, convert DK to K, DT to T, SC to S, KN to N and MN to N. J . KUHNS G.ALTSHULER 2. In the name field, replace multiple letters with a single letter. J.KUHNS G.ALTSHULER 3. Remove vowels, W, H, and Y but keep the first letter in the name field. J.KNS G.ALTSLR 4. The first digit of the code is obtained using PFl*and the first letter of the name field. Remove this letter after coding. J.NS G.LTSLR 1 3 5. Using the last letters of the name, use Table PF3*to obtain the second digit of the code. Use as many let- ters as possible and remove after coding. J.N G.LTSL 16 35 6. The third digit is found using Table PF2*and the first character of the first name. Remove after coding. N LTSL 167 357 7. The fourth digit is found using Table PF2*and the first character of the name field. If no letters remain use zero. After coding remove the letter. TSL 1676 3579 8. The fifth digit is found in the same manner as the fourth using the re- maining characters of the name field, if any. SL 16760 35797 *De fined in Table 2.4-4 19 made to achieve a uniform distribution of codes across a set of logical groups. But in the actual implementation, some phonetically dissimilar symbols were assigned the same Phonetic Frequency code value as shown in Table 2.4-4. Applying the rules shown in Table 2.4-3, the names J. Kuhns and G. Altshuler are reduced to the codes 16760 and 35797, respectively. How- ever, as shown by the groups of letters associated with the same codes in Table 2.4-4, 'T. Vines' would also reduce to 16760, and 'J. Butler' to 35797. Moreover, although these quite dissimilar names would be represented by the same code, 'Kuntz' -- a variant of 'Kuhns' -- would not. Returning to the issue of coding similar names such that they are al- located to the same logical group, it is clear that this principle is especially critical for alternate spellings of the same name. This logic is based on the assumption that non-detectable* errors in record- ing the name of 'Kuhns' may write 'Kuntz' instead. Also, persons changing their name to a more anglicized orthography and having file records listed under both spellings should be more easily traceable if additional phonological-to-orthographic coding rules were introduced. The NYSIIS coding technique introduces a few such rules -- e.g., KN ■* N, Z -> S, as well as a number of ad hoc rules to achieve greater reliabil- ity for the New York State Criminal justice name inventory. However, many of these rules are especially aimed at a large population of Span- ish surnames and may at best not be useful for files with small popula- tions of Spanish surname records. At worst, they may introduce un- desirable ambiguities, as in the case of 'Risque' in Table 2.4-5, which illustrates the application of the NYSIIS coding rules given in Table 2.4-6 to a list composed mainly of OSI employees. Again, 'Kuhns' and 'Kuntz' would result in different codes, whereas the dissimilar Chinese surnames Li, Lu, Low, Liu, Lao would all result in the same code, 'L' . There is obviously still room for improvement, via the introduction of phonological-to-orthographic rules based on linguistic -- rather than ad hoc -- considerations. Moreover, such rules should be developed with a view to accommodating a variety of ethnic surnames used in the United States without biasing the coding techniques such that rules derived to handle one type of surname cause ambiguous coding of other surname types. *This is based on the assumption that most typographical errors can be detected by comparing new names to the existing name list for a file and generating a computer listing of unique occurrences for manual in- spection. This of course assumes an investment of computer and human resources for error detection and correction. 20 Table 2.4-4 Letter Distributions for Phonetic Frequency Code Name Lookup Technique PHONETIC FREQUENCY GROUP CODE VALUE ASSIGNED 1 2 3 4 5 6 7 8 9 S C F A L D E G PF1 Z K Q V P U W B OHM R I N X J T S C F A D M G U E PF2 Z K P B R H N J V L Q X I T W B D F G MRS Z PF3 C K Q V T L P J X N Z E H I U W Y MN TR DRS STN SN PRS STR SR TN TD 21 * co CD u cn -a o CJ lO CO >- CD +J O Q. a. < CD — I C C O •r— -f-J re s- J^ =5 CO I <* CD co UJ Ofr- < - CX < < CD CJ f— 0£ CD f£ co oo S < a: cj < a: CT> s o ex o I— co co S < < cj < ex CO CJ < QC co to I— < ex < ex < CD CO < ex CO ■sC (X < LfJ CJ (X CO < CJ < -a LO >- ex < CO 2: CO CO 1— < < ex < CD 1— CD CO <£ CX I— Li_ u_ CX 1— >■ CSS 2 _J < < CO CO >- < ex < < < CD O X C_) —J ex l— O 2 s: ex CX CJ 1— CO CJ to ex o |NJ UJ CX X CJ LX o CD CD O UJ CX CO =5 O QJ 5- ai CM 00 E 5- O CONSISTENCY FACTOR Is the data easy to record accurately? o ~v "v ^ ~v "^ *-> •~v >* 00 UJ >- ^ ^ "v ^ IN FACTOR i easy to o RECOLLECTIC Is the date recall? oo UJ >- ^ % \> % ^ ^ ^ "^ \> ^ ^ >* RECOGNITION FACTOR Is the meaning of the variable easily understood? o ^ \> ^ 00 UJ >- ^ "V ■^ ■^ ^v %. ^ "v ^ UJ _J UJ < UJ u_ UJ < \~ oo cc i — i u_ UJ - Is the sys- tem capable of handling very long names?* o oo LU >- Are compound names handled? (E.g. are there rules for handlinc surnames (Von/Van der/ de la, etc. ) & place names (Des Moines)?* . I 00 LU >- Are there in- structions for filling out the form with examples showing the intended in- terpretation? o on LU >- Has this infor- mation always been requested in the same way? o oo LU >- IQUE FIERS LU 1 LU 1 Q LU PHYSICAL CHARACTERISTICS NON-UN IDENTI LAST NAME 00 o 1—1 f- 00 I — I cc cc re C_> o I— I OO >- ■o E 03 ■=C Ci_ ZD '..'J O O in * T3 OO i — OO CU Li_i •r— cr <+- a CD 00 CC 1 — 1 +-> >•> >> 1 — r— E E O O CU CU r— — i - ™ -Q -Q 03 03 • U c_) to •1 — *i — -0 1— i— 1 — CL Q. OJ cl a. •1— u A3 CD E 4-> x: CD •p— CU CU s- ■u CU o o s- +J o CU o c_> to +-> Q I C\J CM en Is there an on- going training effort for clerical staff covering this i tem . o z CO LU >- If the data was obtained from extant documents, was a re-abstract- ing check per- formed? o z CO LU >- Was the original form used as the source for data entry? * o z LU >- Was data printed or typed? (not hand- written) o z CO LU >- f Was the form filled out by the individual him- self or by an interviewer in his presence? o z CO LU >- NON-UNIQUE IDENTIFIER LU Z t— co < _i LU < z t— CO u_ LU «=c z LU _l o Q t— 1 X LU CO LU 1— < Q DC 1— o: i— i CO LU O < _1 a. a: h- a: i— i CO CO 1— co -j fO S- fO Q_ CU 00 to o +-> -o a» s_ s- CU 4- 00 c re S- -(-> 00 fO S fC +-> fC -a cu cu u to cu s_ o 4- J£ S- (0 E .*: o cu S- cu +J e 43 "In some systems all information stored originates from the terminal operators. Sometimes the terminal operators work in a fairly casual manner, compared with the card-punch operators and verifiers of batch data processing. Clearly the viability of the concept of building up a data base in this way depends upon whether we can control the accuracy of the terminal input, catching the errors that occur. "On certain systems the errors are cumulative. The files con- tain information about a set of items that is kept for several months and is updated periodically by terminal operator actions. If occasional operator actions cause errors in the files, as the months pass by the files will steadily collect more and more inaccuracies. This situation could clearly bring the system into disrepute, and controls must be devised to prevent it. "A number of factors make real time systems worse than batch systems for the control of accuracy. First, there are likely to be more terminal operators creating the input and they are scattered over many areas. They tend to be more diverse and less controllable than the operators in a keypunch room. Second, the verification operation found in a keypunch room is usually not employed. Third, batch totals and other batch controls often cannot be used, as the transactions originate singly, not in batches. Fourth, equipment failures will be experienced. It is often when a terminal, line, or computer fails, or during the recover period, that errors originate. "We do, however, have one factor that is strongly in our favor in real time systems. This is that in an appropriately designed diaglogue most of the errors made can be caught in re a l time as the operator makes them. The mistake or discrepancy is then rectified on the spot." 4 ^ Because of the differences in approach between batch and interactive systems, the respective weighting factors are being handled separately. (See Form 3.2.3-1 and 3.2.3-2). 3.2.4 Computer Edit Routines . Checks by computer program of value ranges, formats, data consistency, etc., not only have the advantage of allowing for rapid processing of the data, but further act on data that has already been stored in the computer file and is, therefore, past most of the stages at which new errors are introduced. Susan Wooldridge et al, comments on the importance of this process. 44 Martin, James, op.cit. , pp. 70-71 44 C30 c ■i— +■> CT> •r— O) s 'o +J S- =3 +-> Q. E a o •— * c_> x: 4-> u 3 4-> Q. 03 C CO ►—( s- ■o o C M- fO 00 CT> s- c o • ^ +-) >> o CD n3 ^i U_ i-H CO CM a CO E S- o Were check digits used? o CO LU >- Was input batched and were batch totals checked? o CO LU >~ Was data verified as part of the input proce- dure? o 2: CO LU >- LU CC ID LU LU _J < LU _J LU U_ LU i— CO i—i u_ LU LU _l Q X LU CO LU 1— <=c O =n h- c± i—i CQ LU < i D_ DZ t— cm i — i CO CO rD (— <=c 1— CO _i 1— 1 — 1 cm Cl s C i — i i — o CD s- > +J ,, — cr. +J o U CJ fO S- +J cu 3 +j Q_ c SZ i — i i — i S- "O o sz 4- (O in en S- c o 'i — -t- 1 ' >i <_> cu to i^ u_ CXI ro CVI CO 5. o Are there adequate controls over who can change which data items? o 2: 00 LU >- Do the programs allow the operator to see the data before updating it? O 2T OO LU >- Are adequate backup procedures available? O 2: OO LU >- Was data verified as part of the input Drocedure? O 2T • OO LU ■ >- IIQUE FIER LU _J 2: LU _j LU Lu- lu 2: h- 00 a: t— ( LL. LU 2: LU —J Q O HH X LU OO LU h- Q rc 1— 1 LU O < _l Q_ rc f— OC ►— 1 CO 00 1— 00 _l 1— t— t ct: 00 00 LU CrC Q Q 2: O i—i t— ■=c O C_3 O 2: O 1— < 1— O r> Q LU [ PHYSICAL CHARACTERISTICS N0N-UI\ IDENT1 |_ LU 00 s: _i 2: 46 "The first program in any system should be an edit, or a verifi- cation program. All data for a system should go into one simple edit program, if at all possible... The basic principle of the edit program is that it detects all errors that are logically possible to detect, including - or perhaps especially - rare and unlikely ones, for these are the errors which, if allowed to pass into the system cause the most trouble later on. "45 Programmed checks will, of course, reflect the characteristics of the data in the file but may include format checking, (is there a number in the numeric field?), range (is the age given as greater than 99?), internal consistency (do the birthdate and current age agree?), reason- ableness (would someone really have 30 children?), etc. In some cases, an error situation is obvious as in the case of alphabetic data occurring in a numeric field. Where a determination of reasonableness of the data is required, the computer program may not be able to make a final determination. In such case the data must be flagged for further action by the data processing department. Other important factors in determining the likelihood of errors in the file are the length of time for which the computer editing program has been available as well as consideration of the points in time when improvements have been made in its implementation. If only part of the data has been run against the edit program, a first step in data cleaning may well be to run the entire file through the editing routine to assure that all obvious errors have been caught. Form 3.2.4-1 lists some of the principal areas of error detection that can be performed by a computer editing program. The analyst is encour- aged to check the variables in the current file against this list. 3.2.5 Manual Checks for Accuracy . Valuable as other checks on data validity are, there is no substitute for human judgment in reviewing the records. Final determination of the accuracy of a record demands that the printed version of the data be manually checked against the original document. In describing the creation of new files, Wooldridge, Corder and Johnson state that: "...after the data has been transferred to magnetic files the first time, it must be printed out in full. Every item must be checked. This is a time-consuming and tedious job; it is the responsibility of the user department, and plans should be made well in advance for laying on extra staff, authorizing over-time, etc. All discrepancies between the computer files and the manual ones must be checked out. Sometimes the error will be in the old files, in which case it must be corrected from 45 Wooldridge, Susan, et al . , op.cit. , p. 58. 47 Are there checks on internal consistency? (Is reported age consistent with birthdate?) o 21 CO LU 5- Are there reasonable- ness checks? (No married 6-yr. olds etc.) O 21 CO LU >- Is missing data reported? o 21 LU >- Does it check for errors in field for- mats? (Alpha data in nu- meric field, etc.) O 21 CO LU 5- Has all data in the file been run against this edit routine? o 21 CO LU >- Does an adequate edit routine exist for this variable? * o 21 CO LU >- LU CC ZD LU o->— 1 LU _1 LU _J u cu 03 CO CU -O 03 •i— s_ 03 > CD .E +-> S- o 4- O 4- E O E O o 03 E co a cu cu CJ 03 i— i a. o E 4- i— i 48 scratch... it is very dangerous to start live processing with unchecked data, and it should never be authorized . Print-out and checking should continue for the first few cycles of the system; undetected program bugs and input errors are not only likely but extremely likely ." 2 ^ Robert Goldstein, in his Datamation article entitled "The Cost of Privacy" has suggested that, as standard procedures are adopted for reviewing the contents of files for compliance with the Privacy Act, approximately 10% of the records will either be dropped or rechecked in any given year. Under such a procedure, he estimates that: "For a typical system, about 10% of the records would expire each year, and 90% of these would be rechecked rather than dropped. "47 This emphasizes the importance of such manual checks. Richardson and Cunningham have pointed out that audit trails can be kept with each data item in the file in such a way as to enhance the review of edited data. "Reviewing of edited data can benefit considerably from 'data- marking' , i.e., storing a bit-string with every item to indicate whether the item: (i) has been queried without correction; (ii) is to be accepted regardless; (iii) has been corrected manually; (iv) is a candidate for automatic correction later; (v) has already been automatically corrected. "48 An evaluation of the current file must include an analysis of the type of manual checking procedures employed to date. Form 3.2.5-1 gives the reader an opportunity to score the variables in his file with respect to these manual checking procedures. 3.3 SUMMARY OF WEIGHTING FACTORS Form 3.3-1 can be used to determine final weighting factors for each of the standard biographical data elements in the table. The check marks appearing in the NO columns of each of the preceding charts should be totalled for each variable and the results entered in this form. These results are used to determine candidate fields where error and omissions are 1 ikely to occur. 46 Ibid., p. 61 47 Goldstein, Robert C. , "The Cost of Privacy," Datamation , October 1975 p. 65. Richardson, M. and A. D. Cunningham," Vetting of Industrial Survey Questionnaires," Applied Statistics , Vol. 2, No. 1, 1972, p. 54. 49 CO u CD o ro c_ =5 O U c o 4- co <*- O Ll_ CT) c •I — +-> JC cn •i — CU I LO CXI PI o Did the personnel receive special training for this work? o oo LU Are update proced- ures audited by printing and verify- ing the updated record? o OO LU >- Is the printed output checked against original document? o CO LU >- Is the record printed out after storage in file? o OO LU >- IQUE FIER LU 2: LU _i < LU LU OO cc 1— 1 Li_ LU +-> (O O Q fO Li- r— (O en o c •^ •T-» -C J= La- en re •r- s- dJ er) 3 O •^ ■i — T3 +j S- 10 fO r— "O =5 c E ^J OO — 1 CO CO F <~ o CUMULATIVE WEIGHTING FACTOR MANUAL CHECK FACTOR COMPUTER EDITING FACTOR INPUT PROCEDURE FACTOR COLLECTION PROCEDURE FACTOR DATA BASE FACTOR INHERENT CHARAC- TERISTICS DECAY FACTORS NON-UNIQUE IDENTIFIER UJ _J < UJ _J < UJ u_ UJ < l— 00 en u_ UJ s: 2: UJ _i Q X LU OO UJ (— 1— < 1— 00 _l o o 1— 1 < o ZD Q LU PHYSICAL CHARACTERISTICS LAST NAME 51 4.0 FILE VALIDATION 4.1 INTRODUCTION Agencies currently maintaining personal data files will be concerned with the development of procedures for complying with the Privacy Act provisions regarding the relevance, timeliness, and accuracy of data in the individual records. The file validation techniques thus established will also have implications with respect to the evaluation of the level of accuracy maintained in the file over time. Statistics reflecting the number and type of problems detected when selected records are pro- cessed should give a clear picture of improvements in data handling. For purposes of the current discussion, performance of file validation is crucial to formulation of a reasonably accurate estimate of the re- trieval reliability for each of the variables maintained in the records. 4.2 VALIDATION TECHNIQUES Section 3 emphasized the importance of the source document with respect to the ultimate accuracy level of the file. The CUMULATIVE WEIGHTING FACTOR column in Table 3.3-1, Cumulative Weight Factors for Standard Biographical Data Elements, gives the total number of negative points accumulated for each variable during that analysis procedure. If all of those variables have approximately the same final scores, then the information gained is not discriminatory, but rather implies what is to be anticipated in the area of accuracy for the entire file. If, however, a few of the variables have much higher scores than the others, these should automatically be eliminated from consideration as retrieval keys. These values are not as directly quantifiable as error and omission rates, but should instead serve as guidelines in the retrieval key selection process and, incidentally, as indicators regarding future use of these variables. Evaluation of error and omission rates must be accomplished through man- ual checks of the source documents against listings of the records con- tained in the computer file. Trie first decision to be made regarding this procedure involves determination of the number of documents to be checked. A reasonable starting point would be to consider using a small test sample, say 100 records, for this procedure, and then to select a sample size according to the procedure given in Section 5.4. It appears that a determination of error/omission rates would be satisfactory if the standard deviation is less than or equal to .01. The procedure described in Section 5.4 is based on this degree of accuracy. Once the sample size has been established, the selection of the actual records can be made. The most important factor here is that the records must be selected from different physical locations within the data base. This is essential to avoid skewing of the results due to system problems such as the existence of bad tape, overwritten disk blocks, etc. Further, since both the accuracy and the nature of the data may have varied over 52 time, randomness in selection of records is essential to a fair evalua- tion of the state of the data base. The number of physical locations used in a given data base will depend upon its size, but that number should not be less than four if a reasonable sampling is to be achieved. For purposes of recording error and omission counts, a chart similar to that shown in Table 4.2-1 should be created. This chart may also serve as an audit document demonstrating the problems detected, resulting error analysis, problems remaining unresolved, action to be taken, and, finally, indication that the file has been corrected and checked. The analysis results for each record should be kept on a separate sheet for ease of use and also as a precaution against the introduction of confusion be- tween records. 4.3 FILE VALIDATION EXAMPLE An example of the techniques involved in manual evaluation of the ac- curacy of the information contained in the files is given in this sec- tion. This discussion is based on use of a small dummy data file con- sisting of one hundred records constructed through random accessing of the telephone book and the random number generator. Separate accesses were made for each component of each name and for each number group. Thus the entries are entirely fictitious . The ten records (in this case ten percent of the file) selected for validation are shown, in the most accurate source format available, in Table 4.3-1. This data is assumed to exist on the application forms filled out by each of the individuals. It is also assumed for purposes of illustration that the printouts shown in Table 4.3-2 were obtained for each of these records from the computer file. Table 4.3-3 contains the results of the manual check performed on records number 20 and 30. In this case, data has been stored in the file for a set of twins, HECTOR and HARRY GENERAUX. As is clear from the source documents, the twins lived at the same address at the time the applica- tions were filled out. In the intervening period, HECTOR moved to Laurel Street, and a system update was initiated by the operator when he re- ported his new address. The update to the record was made on the basis of SURNAME, and BIRTHDATE, and thus it was easy for the mistake to occur. In record number 70 an obvious case of number transposition has occurred in the address field, in this case affecting both street and house number. Record 80 has a mistake in the spelling of the SURNAME where an ' I ' has incorrectly been input as '£'. And in record 100, a mistake has occurred in entering the birthdate. These errors are summarized in Table 4.3-4. Use of these rates is discussed in Section 5. 53 >- CO •o m S- u o LU u or. 0) o OS ■a ro CD o 14- +-> ra £E o -a LU LISTING CHECKED: BY (initials) ON (date) CORRECTED BY (initials) ON (date) LU 1— O UJ 1— LU O a: CC cc LU INCORRECT DATA MISSING DATA SOURCE DOCUMENT(S) O 1— 1 oc o_ CC I — 1 CQ LU s: LU _l LU =1 1— Q 13 CD o S- <^ S- o o CD cc cd Q. E A3 CO C CD o i/> +-> c CD E u o Q CD U s- Z3 o co I *3- -Q fO CO CO LT> LO CTi CT) cr> C\J O CM IX) t— 1 1 — 1 CO CO cr> 1— 1 cn LO CT) LO CT. 3: #v #* en 1 — 1 t — 1 i— co 00 CM CO . — 1 #* ^ CD CM at 1 — 1 CO CO 1 — 1 CNJ CO ^ CO n CO »— ( CT> CT) *\ CXI CM ■=d- CT) ID CD CO 1 — 1 < — 1 CD I— 1 s- S- >, . — 1 1 — 1 1 — 1 u_ .D. *^ *v JZ5 t- OJ CD c. «* t- ^ o E r^ r^- E 0) jn JD (T3 00 CD 1^- OJ CD -Q E E =3 I — 1 -Q LU 4-> >i >> +■> O CD CD S- O >> h- CD- 1 — ' CL +-> > O JD >> +-> 1 — < CD 13 =3 CD U O CD CD fO O 3 Q CO r "D ■~D CO O ^ Q Li_ s: O ■-D CD c >> +J • 4- CD CO 'O •1 — > CD 1 — CO %. Ll_ ct Z! CD LU +-> >, C S_ o; <+- CD CD CD -a CO CD 3 q X C c 3 c CD 1 — 1 11 > n3 Q E E C C <=3" CD S- ■=£ _l =C CD CD QJ CJ S > CD • S- S- > CD CD CD "O • h- 00 h- (— <: CO 1— 3 CO CJ3 C ^ LU CM LU CM CNJ CT) CO 1 — 1 CM CO O - 1 — 1 LU y— _i cc: O CO LU f— at CD ^ - C_) at CC 1— H CO nr 2: 5- 3: C_> i — i LU LU "-D 2: rs —t or at C_) at _i LU 2: ID at LU LU 2: t — 1 ■=t 1—4 (— < 3: LU CO CO cd O 1 — t _j z: at CO rvj CD =fte LU O O CO 1 — 1 CM CO ^d- LO CD r^ 00 cr> CsJ < 1 — 1 CJ +-> ID O U. CO r— ■ Q. Q. +-> E S- O Ll_ fO CD "O CD CL JZ ZD CO 55 cr: i — i CO CU 00 rO CO ro +-> rU O >! E E CO 3 CO Q UJ oc; Cu Q JC Q +J 5- QJ 4-> CX E o CJ> r j i ro <* QJ DC UJ s: on =« to U LO en LO LO CO 00 cvj O CD LO O OJ CVJ LO co -\ CM CO \^ ^^ CO CO ■^ "^ CO oo LO ^^ o CO" ~\ s „ co i — i Cvj cvj ^v. CO 1—1 1 — 1 r-^ r> . — i o I— 1 CVJ ■=3" I— < o en r>> i-» cr> r— i ■ -H , — 1 CVJ LO 1 — 1 > +J • <4- QJ ■a i — •r- > CU c. oj U_ =£ =3 o s_ >> C m- a» rs 0J ■a jE QJ X E fT3 3 c QJ +J T > o E _j c o c CnI CU S- < a,- CO o 2 — 1 > CU 1- > a> o . QJ CU ■a CO 1— ^ - UJ (— Crl >- o Oi LU UJ < OH O Q Q UJ h- _l cr: s: =c Z O <=c KH LO X s: >- 3 < r-j ^ o s =^c < UJ < o _J Q a: Di ZD 1- ^ CD — 1 —I ^ Oi CO CD M o o o o Q CD o o o CVJ CO =3- LO VO t*v CO CT> o 56 «=C Q CD >- CQ O UJ o o oo Q O UJ LISTING CHECKED: BY (initials) ON (date) . — ■ CORRECTED BY (initials) ON (date) ERROR DETECTED INCORRECT DATA X SHOULD BE 3872 X SHOULD BE TREMNE MISSING DATA SOURCE DOCUMENT(S) APPLICATION FORM X X X X X BIOGRAPHICAL DATA ELEMENT UJ 2: Cf. OO FIRST NAME STREET NUMBER STREET ADDRESS DATE OF BIRTH 57 JC +J 4- o S3 o r— I o (/) (.0 >> (0 c O O) +J cu Q I/O C o •l — to m •r— E O "O c n3 t/i 5- o s_ s- Lu <+- O >) S- 03 E <0 E i— =3 •■— 00 U_ I CO d) r— -Q (0 ! Qjl- lu c_? h- IUJ cC oice: t— * * luiq: 10,000 x error rate x (1 - error rate). Since the product of the error rate and its complement is bounded by 0.25, a sample size of 2,500 will assure a standard deviation of less than 0.01. But a smaller sample size may be used, depending on the magnitude of the error rate. To determine this, select a small test sample, say of 100 records. Estimate the error rate for this test sample. Then a safe determination of an adequate sample size is given by Sample size = 10,000 x estimate of error rate. For example, if the estimate is 0.1, then the sample size should be 1000. Having determined the error and omission rates, the confidence factors (see Section 5.1) can be calculated. The recall ratio for any query in- volving the key fields for which these confidence factors have been developed is simply the product of the confidence factors. 50 Hoel , P. G. , Introduction to Mathematical Statistics , John Wiley and Sons, New York, 1954, pp. 87-88, also pp. 107-108. This is the usual formula for the standard deviation of a mean value. 65 Assuming that the error and omission rates for the variables SURNAME, STREET NUMBER, STREET NAME, and DATE OF BIRTH were 0.01, 0.03, 0.03, 0.01, respectively, then the resulting confidence factors would be: SURNAME 0.99 STREET NUMBER 0,97 STREET NAME 0.97 DATE OF BIRTH 0.99 and a query involving all four of these fields would tend to retrieve 0.99 x 0.97 x 0.97 x 0.99 = 0.922 or 92.2% of the relevant records. Use of these data will be shown below. 5.5 PRECISION AND PROBABLE MATCH Given a certain retrieved record, the question arises: Is it in fact the correct record for the requester? The answer to this must be ex- pressed as a probability. The theoretical formulation of this probabil- ity problem is presented in Appendix II. A computer program was written to calculate probability values resulting from the use of various combinations of keys. Recall that the basic pro- blem is to evaluate the probability that a record will match a requester given the presence in the record of a number of key values K, , ..., K which match the request. The probability values are developed in terms of the following auxiliary quantities. (1) p = the a priori probability that a randomly selected record corresponds to that record desired by a requester. (Notation in Appendix II, Equation (16), p = P(M. )) The range of p should be from 1/T (where T is the largest population from which the data base is selected) to 1/N (where N is the number of records in the data base). For each key j on which the search is conducted, let (2) a. = error rate in key j (a proportion). (3) b. = relative frequency of omission of data in key j. J From (2) and (3), we calculate (4) c. - 1 - aj - bj . Thus c. is the probability 66 PCKjlM.) given in Appendix II, Equation (8b). (5) Let n. be the number of records with specific matching data in key j, i.e., the number of records retrieved matching the re- quested value of key j. Let N be the number of records in the data base, then define r. = n./N Thus r. is the probability P(K.) given in Appendix II, p. 176. j j (6) For each key value j we define s j = c o /r j- (7) We then have from Appendix II, Equation (18), the probability P(M. |K.) given by (8) An auxiliary quantity F (Appendix II, Equation (23)) is then calculated: / ' \n-l n /l - q.\ where n is the number of key values which match the request. The quan- tity F is used as shown below. (9) Finally, the required probability (i.e., the probability that a record is that desired by a requester conditional on a match on n key values in the query) is given by V= l 1 + EF where E is a quantization of a certain independence relation between the keys, (see Appendix II, pp. 179-180). The smaller the value of E, the greater the ability of the keys to discriminate. Independent keys mean that E = 1. If E < 1, then V > 1/(1 + F). The probability value V corresponds to the precision of a retrieval of one record. Results of the computer simulation are shown in Appendix I. The tables should be read as follows: 67 (1) Each table is identified by the data fields used in the simulated query. All simulated queries utilized the sur- name field. (2) The surname decile identifies the discriminatory power of the particular surname used in the query: Surnames in the first decile are the most specific (TREFFTZS, for example) while those in the tenth decile are the more common surnames, such as JONES and SMITH. (3) The column headings reflect the three sample file sizes which were simulated: 1 million, 4 million, and 220 million records. (4) The left column of the table reflects the recall factor* obtained by multiplying together the simulated error rate for each of the fields utilized in the query. These rates were varied in parallel from (no errors) to 6% errors in increments of 0.5%. All fields were assumed to be subject to the same error. (5) The body of each table tabulates the precision of each retrieved result with respect to the query. Values of 1.0 imply that unique records were properly identified. Values less than 1.0 imply that the probability of unique identification is less than 100%. The model reflects the precision and recall resulting from the attempt to locate the record pertaining to a specific individual in a data base of the specified size, subject to given error rates, for several types of hypothetical queries. Three file sizes are postulated: a file of 220 million records, one of 4 million records, and a smaller file con- taining only 1 million records. All the queries are assumed to consist of a surname and one or more other key values. The discriminatory power of a query varies with the discriminatory power of the key and key values used and the confidence factor in the data base. The precision of a retrieval made against the three file sizes given the retrieval confi- dence factor of 92,2% in the example of Section 5.4, and utilizing the four key fields of surname, street number, street name, and date of birth can be read from the fifth line (recall 0.922) of Tables 5a - 5j in Appendix I. The tables present 10 different values for the retrieval precision of such a query, depending on the value of the surname: names such as Jones have far less discriminatory power than names such as Trefftzs or Krcmar. It can be seen that the precision of a query im- proves as the file size decreases. Files smaller than 1 million records will provide proportionately better results. * The definition of recall is given in Section 5.2, 68 6.0 CONCLUSIONS AND RECOMMENDATIONS 6.1 CONCLUSIONS The preceding sections have addressed the problem of accessing individual records from personal data files using non-unique identifiers where accessing includes updating, purging, and other bookkeeping tasks. The value of the legislation requiring Federal agencies to make such records available on request to individuals whose dossiers exist in their files is obvious, but compliance with this legislation is a complex matter for which there can be no single generalized approach. Each agency must determine independently the most appropriate solution given its own operational environment. Exphasis in this document has been on one aspect of the required imple- mentation: the retrieval of the individual's record through use of identifiers that he can readily recall and that do not include his Social Security account number. The analysis has included examination of the characteristics of standard identifiers with emphasis on the varying selectivity of these variables. The high discriminatory value of the NAME variable resulted in inclusion of descriptions and evaluations of several of the currently available name lookup algorithms. Techniques applicable to this type of retrieval problem were discussed along with indexing schemes ranging from the most straightforward to complex systems of encoded pointers. The selected examples for generating the precision tables from the mathe- matical model (see App. I) clearly show that retrieval through use of non-unique identifiers is, indeed, possible. It must be understood, however, that the search keys used in these examples are illustrative only. In practical cases it may be necessary to employ search keys with more discriminating power. Furthermore, such retrieval has been carried out in a variety of ways and has proved to be practical. For example, the 3-level index scheme developed by the Immigration and Naturalization Service presently retrieves 5,000,000 records (anticipated growth up to 10,000,000 records) with a response time per record measured in seconds. The discussion of available techniques also made it clear that the degree of success resulting from use of such techniques depends directly on the quality of the data items being used as search keys. (The assumption is made that the query data itself if correct and that it has been entered properly.) The reliability of retrieval through use of non-unique identifiers was further demonstrated through use of a probability model programmed to generate the precision (i.e., the probability that a record if retrieved as matching the query, as formulated, .on a variety of keys, will be in fact the correct record), given a wide range of recall factors (i.e., the probability that the record, if in the file, will be retrieved based on a range of confidence factors simulating a reasonable level of error and omission rates). Results from this modeling procedure again indicate that retrieval mechanisms based on high discrimination keys are practical . 69 It is also clear, however, that these techniques must be implemented with great care. The use of non-unique identifiers results in very high pre- cision but rather low recall. This effect is obvious from considera- tion of the fact that the more identifiers used, the more complete the identification of the individual becomes. At the same time, the intro- duction of more keys increases the potential for problems in matching exactly the information that is contained in the file. For this reason, a Soft-Match Technique, such as that described in Section 5, may be necessary as back up to the retrieval procedure. 6.2 RECOMMENDATIONS The Office of Management and Budget has interpreted the language of the Privacy Act to mean that the mere existence of personal data items in a file is not sufficient to bring it under these legislative constraints. In accordance with the definition of "system of records" (Privacy Act, Sec. 552a (a) (5)) these items become relevant only when they are used as retrieval keys: "This language further suggests that the Congress did not intend to require that an individual be given access to information which the agency does not retrieve by reference to his or her name or some other identifying particular." A first step in compliance will then be to determine the exact keys cur- rently in use and from that list to determine whether or not the legis- lation applies. The current considerations apply only to the specific problem of the retrieval of an individual's record. But even this single aspect of the legislation will require a review of existing systems to make the following determinations: 1. Are the personal identifiers contained in the file necessary? 2. What non-unique identifiers are currently available as retrieval keys? 3. Are these currently available retrieval keys adequate for accessing the required records in a cost effective way? 4. If not, how can the existing capabilities be expanded or modified to make the required records accessible? Each of these considerations is discussed below. 51 Office of Management and Budget Circular A-108 and accompanying "Guidelines for Implementing Section 3 of the Privacy Act of 1974," Federal Register , 9 July 1975, Vol. 40, No. 32, pp. 28948-28978. 70 6.2.1 Evaluation of Current File Access Capability . If it does apply, the agency should make a determination of acceptability of the current retrieval capability in the context of the legislation and the 0MB Guide- lines. Using the search keys that are both available and relevant, weighting factors should be formed as outlined in Section 3. This pro- cedure should be followed by an analysis of the error and omission rates for these variables as suggested in Section 4. From these rates, confi- dence factors can be formed, using the methodology described in Section 5, and a determination of precision and recall made for some combinations of these variables using the computer-generated tables shown in Appendix I. If the variables that are already available as retrieval keys are not adequate to the task, then current system capabilities will require expansion. Some discussion of available techniques is given in Section 2 for consideration under these circumstances. It should be noted, how- ever, that the Office of Management and Budget has stressed use of exist- ing capabilities whenever possible. "...the development of new retrieval and indexing capa- bilities is not encouraged, rather agencies should ex- ploit existing capabilities to serve individual needs. "52 6.2.2 Evaluation of the Data Base . The most difficult aspect of the analysis procedure is that involving evaluation of the data base itself. An orderly sequence of steps should be taken to evaluate any data base in order to bring it into compliance with the requirements of the Privacy Act. A similar sequence should be followed when establishing any new data bases containing personal data. For existing data bases, the re- quired actions fall into the categories of: 1. Analysis of present data validity, 2. Correction of errors, 3. Retirement of obsolete data, and 4. Establishment of a review cycle to keep data current. For new data bases, the first of these steps is replaced by an analysis of data requirements and initial data collection steps. While it is unrealistic to assume that any data base is error-free in any data field, all fields which comprise the data base should contain data which are as accurate as possible. Error rates which ex- ceed four or five percent tend to throw the utility of the data base into question, and fields with such error rates should be either corrected (at some cost) or deleted from the file. The basic test of any personal data base is whether the people represented by the records in the data base can, in fact, be found and shown to be the individuals to whom the records are assumed to refer. As time passes, addresses change, names change, ages change, occupations change, and people move. M Loc.CitT 71 Files which have not been subjected to ongoing verification and update activities, particularly those in which the subjects are not actively involved in the maintenance process, may be found to contain a large cast of fictional characters. The requirement for the maintenance of such files should be reviewed at the highest levels—certainly such information should not be made the basis of substantive policy decisions, and the \/ery need for the file in question should be carefully considered. Files which are really necessary, however, should be evaluated periodically to determine their accuracy and to correct erroneous data. The utility of such files--and compliance with privacy legislation- requires that a high level of accuracy be maintained, and that suitable resources be expended to insure such a level of accuracy. 72 APPENDIX I PRECISION TABLES Table IDs Variables Selected Pages la - lj Surname and Sex 74 - 83 2a - 2j Surname and Street Number 84 - 93 3a - 3j Surname and Birthdate (month/day) 94-103 4a - 4j Surname and Birthdate (year/month/day) 104 - 113 5a - 5j Surname , Birthdate (year/month/day), Street Name and Street Number 114-123 6a - 6j Surname and State 124-133 7a - 7j Surname , Sex , and Birthdate (month/day) 134 - 143 8a - 8j Surname , Sex , and Birthdate (year/month/day) 144 - 153 9a - 9j ( Surname , Street Name , Street Number , and State . . . . 154 - 163 10a - lOj Surname , Sex , Street Name , Street Number , and Bi rthdate" Ty"ear/month7diy) 164-173 73 en lu Li- ar ' — en U ! UJ I— UJ uj r Qi a U- £ CO i_i X d — u r to to '3: LjJ CD Xj. LU I3X k- hi i lu ijj UJ 1 UJ 1 UJ i UJ i UJ ,j_. 1 IjJ 1 IX I UJ 1 III 0"; ro ITJ r--. "Vj" ffi I.I 1 IV -. M LTj ;t~i 'Tj- O'l * J.* ^* J? ?■? ?■■£ -'•-: rx I X' J •sj- 'X 1 cn U") «-t 03 "KT iS IV TO CS 'X ro en IX) i 'X IX' 'Xi m U i IX ■>x \t "X ro ro CM CM i ro ro ro ro PO ro ro ro ro ro l"<3 ro ro cs 'S is iS a IS CD C3 CS iS 15 i IS S IS iS IX"! IS >S IS' E 'S iS IS 'S i UJ UJ LU IX IX IX LU UJ LU IX IX LJ LU 1 Uj [V CM 'S .33 ■•— : \r X' CS IS IV ■^-i ■r-', 1 IS PO CD r- IV CM iS ro CM 'Xi 'X ■"X IT', i iTi t-i LTi ro T Xi r-- 03 ro ro IT'; 03 I rv 03 cn f-H to IXJ 03 ■r-H m cn ro IV Cj 1 LI: 0~'i ro 1" CM IX CSJ IXJ ex ro 03 C'-J IV 1 "^ CI f-H cn QO U3 m ro ~-» CS 03 r-- in 1 03 03 03 jV iv rv IV r- IV IV 'X 'X 'X 1 »-■ —■ ^— ' — ■ " ■—i '""' — i *-' ■: — ! ■•~ J - r * ■r^ _" X ' .X ^J X, x IS " ^_" ^1 'S Ijj I (S iS CO IS iS LS IS CD iS IS 'S' iS ( \-i >x - L - "^ ■* — ■^ -" — ' -'- ■ l! - ^ - 1 -* 1 LU LU IX LX Uj IX ijj ijj UJ UJ IX IX IX i iv ro IS 33 Ul ro (Tji IX CJ 03 IX cn 03 : is ro U3 'X 'X cn ■sf ~i ro CS ro ro CJ i CM en iS in 'X IV iX rv ro ■f o '' 'Xi -x i S ■^-i CM IS IV CM 'X 03 cn 03 m ■<—* 'X 1 CM en 'X ro Oi •X iM 03 \r IS tx C'J IV 1 03 tt ■»-H 03 T *-H 03 \r t-H 03 "X ■*-t r-- i IV IV IV 'X 'X 'X IX IX IX "X ■vT ■x ro i IX) m 10 IX' IX) in uo X.i Xi in in in in 1 IS 'S iS 'S 'S IS »3j iS iS iS IS 1 E> Q CM ■tf 'X '71 CM 'X in IS' ijj i O •s o IS iS S •i--- *~H CM CI ro ro 1 iS CTi 03 iv 'X Xi -tf ro CM ■^-t iS cn 03 i is CTl Ul 'Xi cn 01 LTi Xi cn o-i 0'1 03 03 iS iS iS IS iS iS IS 'S iS iS 'S 1 1 25 vs M >t ^ V* i-i 3? J* K Jl* 1 a iS IS IS iS IS iS iS IS IS iS iVl IS 1 iS in iS Xj iS in IS IX iS Xi iS in CS iS IS ■•-J CM ro ro •^r ^ X in 1.0 75 Ul a h- O u. CO I- _J if} UJ on UJ LU UJ LL. u- O 01 : i X jj > — l !D 2 . CO in •:X <—■ UJ Li.' G_ O) IJJ •X a. (K a a Li. LU m ro ro ro 10 ro I s J i'o r- _ ro ro 10 S3 'XI CD i..j CD CD '3 UJ dj Gj ijl,' CD I 1" 1 1 IJJ i LU I l-Ll 1 UJ ill 1 IJJ LU 1 UJ i : j 1 U.I in LG T-i 01 r-- ro '■X: u i 0"'; ■T— 4 CD CD in Ul ro ' r -t r- is? CM LO 1. !~l a-i ~-i T$ -""*- |"-n E' 10 I'O T-. Ul UJ \j" CT) CM ■r-> 02 CO ro CM '■£■ ID CO 1. ;". ': i. r< iJl i, n CD l n ro CD IV Ul !'"0 CM ■r-l ■r~: i"'.i rn 10 \y Ul Ll U2 IV i 1 i iJl |VJ ■r-' CM t ro CM •*-• CD iT) 01 IV •■jO 'jj Ul -.1 i Ul CTl 01 Ul 03 IJJ IJJ 03 03 CO o: CD CD CD CD CD i3 CD CD CD '3 CD iS cs CD C ID ID UJ r-- LO ro Q i a i *—i \o 13 LO rH in rvi CD ui ,-H i.n CD kO |V| iTi L 1 LD t- a i_£ •ST; 13 13 J4 O ■3 1 ID i ID 1 ID i UJ ro Ul j--. on r : LO IX) [■'- '•,;" ■»— i "tf- 1 I ; i.n in IV 01 CM I.O «— 1 LD CM 'X 1 01 r-- iV '- ! CD UJ cp f ID UJ ID UJ ID .: j. CL xr I. -1 . i '3 "■•.j Ul OJ ro ■3 LD 111 •^-H Ul i'J I Iv I ■.! Q3 ro UJ LD Lll Ul Ul Xj" XT \r xr \j* Xf '3 CD c. CD i3 iVi 1 CD' '3 |3 '3 IX' '3 '3 3 13 13 IX.I S3 ■3 '3 i3 13 13 'X' 13 '3 '3 '3 •3 '3 III ID LD ID LD LD ID ID ID ID ID UJ ID ro Ul <3 IJl OJ IJl ro Ul IN- '3 01 ■^H 01 -t 'Jl '3 Ul ro tj- •j- 1 1 f OJ CM r_ i3 ui i . i XT -s LD CM *-« ro 01 IJ'I CM 01 0"'i CM in ■vT ro CM CM OJ OJ OJ ro LO CD ill *— i m ■^1- 01 CM l.O '3 xj" 01 CM LD '3 ■vT CTi oi I.O ^r ro ■*-> '3 01 '■D i.n ro CM <■ UJ ,'"1 01 01 ni 01 01 r- r-~ 1'-- IV r- IV LD *~ l ''~~ , " ,— * "~* '""' ' I— ' r-l ■ ,— ' "~ ' — ' 11-5 ■'"^ "^~* 13 01 ITi IV Ljl Ul IT, '3 i3 Ul Ul 1T1 1: 13 L 3 C ip LO ro ro en 01 65 13 i'-. LO U : 13 76 — I Lu - LC' O'i *-■ i'J"i v" Lj"": [■0 Q i r- iv. Cn CV! cu \j' ro r-i ; i n r-- \l -^-; cn CO ro ^ — i CT'i r •- i,-i CD r- •ct CD [\_ r-- .,.-, PO cm CM CM i— i *—* -■— ; CQ «3j CD !"0 ro FO ro ro ro ro ro ro it CD a CD CD © CD '3 S3 S3 S3 UJ LU LU p^_ FO ro 00 r- cd \i CVI CT» CV! ro \i ■•-a ro m cm i^j Oj :3j CD N- 13 LU 'S3 Lf3 G3 153 S3 '53 (53 CD 153 CD CD CD CO 15"' O LO CD CD CD (53 in a in CD CD CD U"J CD LO \T 77 CVI ! LO ! CD CD '53 '53 CD CD UJ LU 1 1 1 LU UJ LU IJJ LU ! t-H ro l\_ IV! CO ro |N"i r-- i CO CD IV CD I>- CD 03 1—3 ! QO 0*1 tv m 'D LO r- CTi 1 *tf IT| LO cm CD 1 1 1 i - r- i CM XT r^- jVi ro LO 03 — - ! ro •— * CTi iTl CD \r CM — ' I (V r-_ CO 'O U3 CD 'X' CD 1 ! i I 1 CD '51' QJ CD (53 CD 'j^' CD CD CD CD CD CD 1 LU LU LU LU LU LU LU IU LU LU UJ UJ LU 1 CV! in evi CM ■■— < in ij"i Tt CD cn ro *'T r\- i LO ro r~ CD ID r>- — i ^— i r\ iTl IT": CO r--- i r- CO **-; •>— * CD *-< h— ) rv- cn CD \T j">.. I"-- i CVI CVI m 'D IV. CO -^H m "\t LO Oi LO tj- | ■»-H ^r IV- ^— i ^r 03 ro k eg r- CVI Q i 'Tt 1 cn i>4 in cn CVI in 'ji CVI U3 cn ro CD 153 1 '5!' CD cn rrt CD r- •D CD in xr t ro ro i |N- r-- LO CD '-"' L0 CD ■D U3 <-V CD CD CD 1 CD CD ^ CV! ■■^J" i.n ITl OJ 'D CD in 1\! Cu CO CD eg CD CD CD t— ! ^-» C'J CM ro CD cn CD r-._ U3 m ^ ro CVI i5Ti ITl CD cn Ui cn cn ui Ui cn cn cn cn 03 n (S3 ! CD 1 CD ! CD LU I — i Ll_ j." ■T. ll i f Ll to l- jj en LU Di os: h UJ UJ S~ tk: •cr 71 IX. a: a ■ o V 1^5 l-H IjJ ^L CO CO ,j 1 — 1 u UJ 0* :x CD a> jO LT) UJ _J r_j UJ UJ ^r i UJ ro UJ J ro i IX! CD CD CM ro ro ro 1 UJ i UJ 1 UJ 1 UJ UJ 1 UJ 1 UJ i UJ i UJ UJ 1 UJ 1 UJ UJ 'J3 CD CO LO UJ |Tj CD m ■^ 03 in ro ro |J.| in CO UJ CO LO in IS en CM o CM 03 10 01 CD 133 m CM ,, r CO ro |V CD CM i. 'j Ij .' ro C'i LI i ro ro ID UJ IV 03 Ui ■r-l CM ■^ UJ CO I'D CM -J UJ CO UJ ■^J" CM T--I Ui |V in ro CM iVt 03 U3 UJ CD iTl CD 03 I--- r- IV IV IV IV IJJ ij3 ■i 1 ' ,— ' ■.-H •r-H .. — i .^-i """ ,r ~* •r— 1 '""*" T-H 1-1 >33 133 03 U3 CD CM CD JJ IJJ vi rr CD CD CD CTi CT. I UJ CT'i r-w iXi r.- 03 iTi CD i. SI CD UJ in i UJ LO ro en m rvl CM CM rv '31' i^j 13 I'l UJ LU IX r\| in IV r*" CD 1 Ul i2i in ijj in ix CD IO *-H en ■r--( T r>- •-" r\i CM in iTl iT. S3 CD CD CO "33 CD CD CD IS 133 IS UJ LU LXl Ixl LU Ixl UJ LU UJ UJ LU UJ UJ LD Tt ro --* i-j IT, 03 IV IJJ in i S iS CD in CD in .33 CD ■— ■—' IS iJj S in CM IM iS CD ro CD m CD in in u i m iS 78 ro i LU i Go lo ro r- n 7 T LU LU ro ro 1 LU 1 1 UJ I UJ i Ul '\T rC 0"'! m \f !'-. :"- ro r- ' V ~!i r-- i.Tl IV- i. ri lj-i ro CM i?..! CM CM CM LU i cm i LU a o I c I i rj i ■— - I _! T'4 CM CM CM cm CM CM OJ C'-i CVI OJ CM CM ! CD G5 CD G) 'X! IV - ; ■2D i3D GJ G3 *X ! CD CD f 1 LU Hi 1 LU 1 111 LU LU I LU UJ 1 } I LU ! UD LD ro CO ro LD Rs- iT; •J J ro '£' CD CM i PO [■■_ CD !j~i en IT, [>-■ 01' CO xf h- IV ro i LO en fv_ 0"! T* ro i. o f\; OJ Uj ro "\r CT- ' { -jd *-* [\_ ro CD r- "^T I 'X? ijj r>- ■j~' Ul ! t-H nT LQ 0"l CM xr r>- iT"i |M in iXJ —i \i ! LO ^ ID p.j CM t-* CD C£l Ul CD r- !V r J_! i i\- N r^ r- r^ iV T- r*- tjD '- j co i£i Ijj 1 G) G7 IS CD 'X' CD Gi GD K3 OCi CD E CD I CO !— _J 3 en LU LV « LU > LU pi 'jj IjJ I ! U I CD CD '31* CD 'X- !_^l 63 i\_i '3Z' Gi ! UJ LLi LU ;,. LU I . 1 UJ IjJ LU UJ LU LU IjJ I UJ -■— i pi- UD IJ 1 |\- en 1 1 1 in I'J -.— ! r-l '.0 1 lq Tj" ro TT in ■— I 00 CD —■ ■ !'-- '-£' 1 1 1 in i C'- G "; CM n t f- CD in \j" i'- C'-J "—5 ro U 1 ' C CD CM *J CO ■\j" Ei OD rv- on i3~! ro r- i ro \t LO ID N- :-T: *— 1 P-.i st 'Xi iTi ■*—< ro i r-- \1' -— CD in CM GD r-- '^r .--, 01' Xj (T3 IX a a ■:T X o UJ X L£ a Di LL LU '.-■ 1 CM CO CM CM C I OJ CM CM cm 0-J CM CM CO a CD Q CD iS CD CD co 13 CD Ei 'Xi LU LU LU LU LU UJ IX IX LU IX IX IX UJ CO CD Xi 9 CO IX ro ro IV •i-* ro rv •>~1 P0 CD ITi ro •X ro Ul I-* CM IV CO f-i iv *T 01 Ul C9 •sr 01 ■~-l ^ CO r- 01 Cm K- LO - ■"ff id 'Xi r>- |J: CO CM ^r Ul 03 ■^-t ro 'X' T/ CO CO 'M CD Li i — i r-- ro 0") 'X CM 01 0. Ul on CD IV r- IV CO 10 IJ1 LO Ul ■*■}■ ro n ID ro ro ro ro ro ro ro ro ro ro iD ■>-< IS CO © S3 CD iS CD CD IS CD CO CD T— 1 CO CO CD CD CO i3Z' 'Jl' CO CO co IS CO Ul UJ LU UJ UJ IX.' IX UJ LU IX UJ UJ III Oi -•—I 1" ^^ Ui '2 '.Jl I s i CO IV u ^ in 01 •^r CM Ul CO •r-i "xj 1 CM xr Ul Ui ro i—i CM |JD IV LTi *-i T tf CM IV Ui Cm IV CM ijj IV tr I'M iS? Ui "Jl Ui iT: CM IJ1 Ui CO ro IV IVI r^ CM 'X *-* CD iM IV CM IV 'Jl ui ro CM CO m IV 'D ^r ro rH iS m in in Lfi U~i LO M- ^ ^r ^r "3" •sr ■*$ ro *■"• * —i ~* 1 ~*' ■*~ ' -r-H co 'X j co CO CO iS CO iS CO CD ;>» J>t }£ >£ a ?■£ j^j IS CD CO IS iS iS IS IS U< s IXi CO ID IS CO ?>: }■•: JJ IS IS IS iD Ul CO ro in co CO IS ~< IM IT CD cn CM CD CO ID "S CD CO IS iD iS S3 IS iS ■»-• i— CM CM rn ro CO U) CD IV 'Xi Ul ■"vT m CM ■— < CD cn CD CO Ul Ui en Cm- Ul Ui cn Ui Ul cn 01 CD >S 'S IS 'S 80 \r CM CM IS IS G3 iS IS '3D CD iS S3 •S i<; i-iJ I.Li UJ UJ UJ UJ ijj Li J UJ UJ UJ UJ iS |V kO QD in ■*— ' PO iv ro U3 li ko ID iVl LO Li" U. 03 •— ' 03 iji LO ir.i IJ , CO ■r~1 00 In ■cf ■^-i CO LP IS in IS ^1" ;\_ iVi ro »-H 10 1." I^J \T ili.' IV j UJ CVI CM CM CM ro ro ro ro ro ro ro ro pi; CD iVi CD IS IS IS iS S3 13 Gj I'l UJ UJ UJ UJ UJ UJ UJ UJ UJ LU UJ III f 'X' ro 03 LP! — « CO — i ro IV O'l "T ■^H kO ro i— i UJ U.i cu 01 'jj 03 i u V IV |V ro in CM ■*-H ro nD **H 03 kO i£< UJ •r-i 03 LO CM CO 'I -■— ' ro •»— I ro •^-J T CM kO 'X' kQ kO kO 03 iv kO kO 1. 1 1 iv UJ 'S 0\l ro 0.1 ~-i r. i in LO LD U~i lh m ! 1 1 kO in ro 'S is (Ti iT) 03 N- 1 J_l LTD <3~ ro CM -r— » "-* ""' ""■" ""i CO Ol U 1 CO 0", Ui CTi CTi (S3 S3 co h- _i CD UJ Q£ P UJ LU Ck F UJ •-1 U- Oi. o s CO z: O X Q "llIZ co cn s IV iS kC CM U 1 IV 'S 03 IV CO ^-i ■J. •s ro 10 ■•-> I-- ro ro i.'.j ro 13 ro -. — i 1. Q J.J— 1 i.'-i •^r ■■— • iTi iv CO L~ ITi i,n in CM p. I en i'l ! r._ IX 1'": iS 1T1 03 U3 m '"v" ro CM CM CM 13 '3 i3j J LX Ti (TV --H "T i\> 1"-- C3 ro C . 1 'SI 1 \1- IJJ 13 CM 03 I 1 i "ID CD Xr 03 in e M, n. IJ3 ro ■r-H CD ijj Ui ui m r-H uo a CM 'X a \r CD CM If; Ui r-"i ro ro CM ■^ ■*—> iS '3 (Xl CO CD [\_ l.J"; i.n 'Xi LD 'X IX IX 'X 'X in m m L~ UO :.n Cl> lS '3 •3 i5D G3 13 iS >3 Cjli i3Z= '3 13 a Ll CD h- 03 UJ Dr. UJ UJ U. IX a z: ii _j if' en UJ _i UJ CM CM CM CM CM CM CM CM CM OJ CM CM rj '3 3 O '3 IS Ij..' '3 'JJ IS '3 iS III IX UJ Lu IX IJJ IX IX UJ IX LU UJ uj i"M LO 1 r-- f-H 'X iTi 03 ro 03 Iv" I i3I< '.0 CM 'JJ ■0" m CM in •^r '3 ro *-< r>- CD I: r-- CD T-* r- 'Jj r- '*-* 03 r- Ui ro CD '3 ro 'J'i r>- u"i in 'X en CM !'■- 10 •>— i CD (3 ]3 ■^r m ^r ij» T en in IS 'X CM CD "<* L r ' CM 'Tl p- "U CM CM cm e-j CM CM OJ '3 '3 iS '3 '3 '3 ex '3 '3 '3 '3 ."S '3 CD X ft i-"llJZ co oi LU ex I- LU Q£ LL O a lu r x a: — ■ UJ z co 01 -a: 1 — 1 <_> LU a: LL O) rt3 in in in LO in in in m in in in in in CD GO cs iS iS CD go go Ei CD 1 1 1 1 1 1 1 1 1 1 i 1 1 1 UJ LL! LU LU UJ UJ UJ LU UJ UJ UJ UJ LU 1 — 1 0*1 |V 01 ^r \r \r CTi CO IV CM •r-l CM m 1 xt r^ t VD CM CM 'X in Oj Xi 03 \r ■st 1 f-V ro IV V in U) Xi IV *~H cn IS in 10 ■sr VD cn *vT © [\- Xi Xi 0.1 CS in CD iv jr XI cn CVi VD 135 ro 1"- t-H in Si -vr Ci ro CO LP ro CD 'Xi in CM CS 1- LO CM 0*1 iv C3 O h- _i UJ CO 2: LU o ai £3 — F •y 00 60 ■X »— t u LU Oi CM OJ LU UJ U- •i u O u CK U.I 'T CD i3Z> •3 i3 3 i!5D OJ IS i3 '3 '3 i^>"i 133 13 S3 '3 '3 IS 3 '3 3 '3 in LU IjJ LU UJ UJ III UJ UJ UJ UJ IV CTi CO ro m ro LD n— . '3 CM l£i pTi IJD S3 i-H CO CM CM en CM *-H IX' CM •*H '3 CO in r<~< i_jj LO in CTi '*\t ro ■^-t 'T'i IJD ^r On! '3 r-.._ U i CM (3 ■J, CTi CO CO CO Uj OD IV iv r- |V iv. r>- IV IV r>_ |V [V iv IV i\_ r-_ >Tl en Ul cn iji 1T1 IT. lT: iT'i 01 iTi CTj iji iTi en CTi o\ CTi iTi iTi U', |J', i3j i!D »X< Ej CD i3L.- LU LU LU (S3 tt "S3 O CD S3 S3 CD c r"i l 77- ! G) »S GD ■53 iT - ! IS GD ■x» iS '£) '3 1- iS CD IT. 01 UJ C3 LU 13 13 i3 3 13 i3 13 3 '3 '3 13 IS LU UJ UJ UJ UJ LU LU 13 IS 13 IS S3 uJ ij 3 i3 — CM \T IJD 3 IS (S iv ijd in ITi iTi iTi CM IT', '3 iS m is 13 in IS m cm ro iS LTi IS i3 is in 84 i3 '3 IS G3 is in — i 13 I G) I ro ro iji OD 03 CO S3 S3 '31' CD C3 © i i_l i LU UJ U.I J i-> ! L^i CO •*— .• I _l 1 Uj 'Xi CD ! ! CO ) \ i |1j i i^-. | h i S i m P- i IS 1 Xj" ^1 vr ! CM 1 CTi G~": iTt 1 '. '\! ! - CO CD CO UJ UJ CO CD 03 •3 |\- IS CM '•i i U.i \T \t CTi CTi ll.l UJ UJ hi " CTi r- .■— ' CD r-- '•U UJ ■*— i 0""! CD ro Tt l^~J T-- U3 -,-1 lo l!"l "^t \r pn \T V f \r on 01 CTi on ■4 -CM o"'i lU o UJ r-j en UJ S CO CD CO CO l~l iS -*- — ' -^' — , - j - UJ UJ Hi II 1 LU lu Q . T— ' CT. t~; 0"' CD "^J" 03 CH CT"; IX LD £1 IS 1T1 ^— > 1\ I CD 133 CTi CD IV UJ LO ■»— < CD CD co CD PCs CH cn U'l CTi CTi 1T1 CT' cn 03 CTi en CTi CO CO iS CO LU rn f-n p-j j r-- LO ! r- Uj i <3 cn 1 'S CTi 1 CO h- UJ LU CK UJ Qi LU I t— J UJ ! UJ CK 3 CC ! Q£ i_i 1 LU Cfj I z; lu i o a; o 1 - H-C ! CO 03 -I 1 0-i iS CD 13 CD CD >S 1 2: ! LU LU LU Lu LU UI 1 O ! 03 CM 0-J CD ro xr 1 — 1 IJ3 U0 KD '•D in M ! _l 1 LD ro ■«— ! CTi r-- in 1 _J I m LO 03 "^r xr •xT i I-" 1 1 1 1 CD m UJ 1 I 1 1 1 1 1 21 i IT, CTi cn ij-i CTi CTi i 1 cn cn CTi ui CTi CTi : —1 i UI 03 CTi cn CTi CTi cn cn CD CD CD 'S LU CD iS iS iS UJ UJ LU i •J\ ro 1 CM F-- 1 CT. 1 Q'l 03 CO 1 CTi IJ' CT'i 1 l_J UJ Dl 0- I J2 CO O) re C_J J O) t — _Q fO ro uj -i cj ui « ■'■X a CD S CD CD ea iS iS iS Si iS iS li.1 CD © CD © »S a CD CD CD S CD iS IS s HI l-U LU LU UJ UJ LU LU UJ UJ UJ UJ III ro ro Ul r- "T UJ iv CM 'XI CD CM ro •^T in CM S co ro CM I'O T ■,— * ro KD CO U3 !'-.„ uj iD3 CM IS iS CM ij - i CM iTi |V s I.X) IV CM in |'-v i ■■- Ul ■r-H LTi CO CCi |V Ul —i iv CM IV CM IV CM ID CD ■tf 03 CM i.n \r> ro CM CD U, iv ijO ^ 10 »-H 01 CD un s\\ (VI CM CM •»— i ■-—< i — i ■F— l T-< T-« iS CD 03 03 03 ro 03 CD CO CD CO CO 03 CO 03 iS © IS 'X' IS CD 'S Si iS CD iS 'S IS iS CD iS CD iS CD s IS i* - : 'S 'S iS iS IS CD 'S i-j " J_ ■* ■* "* "■*" _,.. l„.J li.1 III LU UJ LU UJ UJ IX! UJ !.LI UJ UJ UJ III 1/1 ij3 en PO |V CD IS 'Xi 03 CM Ul 'X' ro ro ro v UD 03 Ln i n 03 LO 'O en 'X' ui Ul •»— CM ro ro ro CM CD 03 uo ■,—i r- cm •JD ro iji LO ■r— 1 IV. ro IT| ■tf iS 'X ■r—i iv CM CM .*-t *-^ -■"" 'S CD Ui U'i Ul 03 03 IV 'Xi 'Xi 'J3 kO i.D 'O U3 LO LO LO i.n 10 \n Ul Ui 01 iT'l Ui 0"i Ui cn Ul |T| Ui Ul Ui Ul U'l ijl Ui Ui u ij'i Ui Ul O'l Ul Ui O'l 'S iS i"i iS CD iS Cj o iS iS CD IS iVl iS 'S iS iS Ui CD '33 iS s 'S •s iS s iS IjJ UJ UJ UJ UJ UJ UJ UJ UJ UJ UJ UJ III '.ll CM •tf 1— ' "tf -^-i \f 'S T-l U3 CM ro I.l3 iS IV •*r Si ,r 7t" r-- CO IS 'X CM IV IS 1-* in 'X 1 i"'- 03 03 03 03 i ; i 03 03 IV IV '■n \T ro CM -<— i IS Ui 03 FV LO in "T ro CM ■»— < *-i *-< '— ' ■•— i s iS CD IS CD IS s iS Ul Ul Ui Ui Ul Ul Ul Ul en en Ui en Ul Ul Ui Ul Ui Ul Ul en Ui Ul en IJl Ul cn Ul en Ul Ul Ul Ul Ul Ui Ul Ul en Ul Ul S CD CD CD a i ■x. £3 'S ITi 'S — ! — ■ CM IV U3 LO xr ro CM ■^-i cn Ui Ui iTl cn U Ui s 'S s iS CD Cj CD ro -r iS X' J4 34 y* JS 34 3i 34 3--' 34 >i 3C 84 IS s iS CD !".' IS 'S iS 'S S3 S IS -iS iS m iS LO '-' in CD in S' LO ' .' in i33 86 i r-j LU ! _! i 11 I CD (3 CD CD CD CD LU UJ LU T— 1 C\l CD in ro \] r-~' p. ■ m T en CM ID CM tji in I'O CD CM Cvj 1 ".! ID ID ID 1.1 .1 UJ rn T 1 ro Lfi i2 U3 ro — . LU ijj L~" LU |-r: UJ IS ID i.n ID IS t— i 03 (jj m CM UJ ro m j'-H — i IV pj n ■rt 03 ij3 CD CD IT) Ui '■-0 i. n LH L 1 cs Ei CD UJ CD CD CD 'Xi GD CD '— CD CD LU !_L_! LU LU LU LU LJ LU LU UJ hi CD IT..] \J" I'J m m 0" El '"■J CTi CM Cm 'S ■5 •^ r— i ^r CD ro CO LO LO \! ro CM ••— < CTi rv in m CM *-i CD 1 '"! iTi ID j p en iJi |— | LTl CD on 03 |-|"! !~l 1 iZD CO 03 CD GO rn 03 133 tjCi Ul Ijl Ui U'i cn Ui iTi LTi cn CD iS CD CD 'S S' iS rr\ CD LU UJ CO <3 UJ in iS CO IV CM U3 U; in CO Iv f- LU CO LU LU ( at uj i LY ! H- IS LU 1 ! | Uj il i_ 1 U"J ! ct s - ■^r 1—4 03 m CM iTi \r ro ro m ro cm CM CM i-i •>-< — i IS i- IV r-- |V IV =■>_ r- rv r-- IV (V IV Q 1 Ui Ul iji iji •'Tl CTi cn Ui IT, IJi ITi CTi Ul CTi ij i en iJi Ui en Ui Ui Ui O'i CD iS CD S CD CD CD ■s CD s 'S CD LU Ql 0_ I Or: I PS C\J GJ LU i - ^r CO in en p i ro ro CM CD IV ro GO ro CD ro rv *-« CO CD ^r on CVJ in LO CM CD |V in rj CD IV in CVJ ui IV ■*vf GO CO CD IV r^ r_ f^- co LO CD in in in Tl id- \J- ■*t ■yr -r ^ ■V-J- \r *tr tj- T Tt G Ei Ei CO _! U_l Ll cc ! - cc o U h _l El CO UJ UJ •— •■ ac QL UJ H- 03 UJ UJtE Qi Ei 'X 2 ~ZL u_ ct: oh =) UJ CO z: ijj o i> a i-h- 2: 00 CO ro I — I u LU a iX OJ OJ 10 Ijj ,J CJ Ui LU E UJ W CO UJ a h- u ■I u UJ ro CD El CD CD ro CD CD CD Ei CD Ei Ei CD CD CD CD CD CD CD CD Ei Ei ro Ei ro IS UJ UI UJ UJ UJ LU UJ UJ LU UJ UJ UJ LU -■--' in CM CO IV ro Ei ro m CD T-) CM ro ■"sT «— * f-i CM in en ro ro uj ro rv OJ ro ro OJ ro ■—i "Xi CD ro ^r r\_ IV ■^ 00 oo CM "3 m 'S3 in U3 CO ro in ^r ro ■*— 1 IT, OJ CD CD CD '■xT i"' 1 CD ro ■j_» rr CM CD r- ■■-t ~H E' CD CD CD ro ro CTl Ij'l ro ro CD ro GO OJ Ei CD CO ro iv IV r>- [V IV r- IT. ro Ui Oi ro cn ro ro U\ ro ro co ro ro ro ro ro ro ro CD CD CD E o ro E ro a ro Ct a ti- ll. UJ N X CD CD iD in E CD in ro ro — — oj <>*. .>* *t >t j.? is ro E ;r> G CO - LU — i lK Di LU I— 03 Ui UJII ck s ■! -T T. U. lii Ol-J UJ 03 ZC LU O L3£ O -- I- z: en en T< Ui OH 03 PO UJ T Ti- \i ro ro ro ro ro ro ro ro CO iS iS s iS iS CO iS 'S ~^ CM ■tf 'Xi IT; r~ iS >3 iS 'S IS iS CD Ul 03 r- CO in •Tt ■s en en u> >Xi en iTi CM CD iS IJ"' iS — ■ — CM cm ro ro cm - 1 is en iTi iTi Ui iTi CD — IS iS iS Ul Ijj 'D ro 'S iV: CO (53 i ■-- 1 iS 'S iS iS CO IS iS co iS C5D iS 'S s iS iS iS iS 'S iS CD iS LU LU LU LU LU Ul LU III U! Ul Ui UJ LU ■«-« — i CM K~ --H in 00 OJ IT'! OJ m CD .p-H — H ^ iS Oj IS 00 oo 03 in ■»— i ro CO ro ^r ■•—• in M" iS O! ITI T "vi iS t-H Uj iS ro \~~i CM l.O IS ro CD CD iS OJ ro roi *\f ro l_i_: 03 in ro S |V ■r+- OJ iTi '■J-' ro co -\r -sr ro ro ro ro iM CM CM ■r-i -."* •r-1 T-H rs_ IV rv iv rv r-- iv r-- l--_ iv r- r - r-- fJi iTi cn tn in iXi Q i iT'i CTl U 1 Ui en Ui 'S iS CO •s s 'S iS © iS iS r* . 'S iS 'S 'S 'S IS 'S s co 'X' iS 'S iS IS iS LU LU LU LU ID ID ID UJ UJ LU ID ID LU in CD U\ 0J OJ pv. CD '-£' m n-H *-H iM ro IV UO in LD ■sr CM 'S I-- ro IJJ ■»-l CM -— » tTi T-i- m ^— < ro T7" ■^r i* , --j iS ID 0J i o en |N- •-H Tf CD *H \r r,_ 'Hi ro in Oj iS CM LO LO M" ro ro CM i~t ^-) 'S iTi U' 03 IV ro ro ro ro ro ro ro r j ~! ro OJ OJ o-J OJ en iTi iTi en Ui U i IT". Ui U: Ui ITi Ui en iTi Ui Ui Ui Ui Ui "Jl Ui Ui Ui Ui Ui Ui *2J »3~« G3 ^r JJ H «•£ 5-i a t-r '>'' }->: 5-'. ;•-? 'C M J! Qi iS Ci i^. 1 iS iS |T"| •S iS 'S iS iS ro 'S CE- in 'S LO iS in S3 in i~ in s U3 iS OC i t : j £V UJ q: Km lll ■ lu i_. .m U) oi n ■ L ^_ .*L_ Ll fl£ O h- ._j LU iT) z: ui UJ DC ■-r CD 01 X 1—4 UJ Q£ 0_ cr cvi O) .-c _J ■a: u LU a: ^r 13 i3 13 13 '3 iVl o IS 13 13 CD CD rn: i3 •^ -"■' - 1 - "^ '-•- -*- "*■ — — -*■ U_l UJ UJ UJ UJ UJ UJ UJ UJ IU UJ UJ hi LO ijl ro |T"| Ul '3 iTl ro IV* "'■•J ^■T in IV. 0J ■r-H ij:i 'Xi TO LU CD CM ro m Oj ro ^-1 UJ t-H i.n Ul ro 'X l3 ro rjj lT"j CM 'X '3 LO -»-H CD •*— i IV pi 1 1 1 ro I'M ro IT'I ■*£*■ CD in sj" M Lfl u i 'Xi 'X' IV r- 1 ! 1 CD (Jl '3 -i- CM CD CD CD ■sf CM i3 03 CO ^r CM •-— i cd ID CD m in Ul m : n ■'i \r ■* '■tf Xf CM CM C\J cm T'-J CM i.'j CM CM CM C 1 ! CM Pil CD '3 '3 LiJ ii.i CD '3 UJ UJ i.n f\J i~M t— 4 ,-,-, in i\l ^— 1 U0 Ln Ui ui UJ IV CTi ^— > 01 on i-? ro m r- i '-( ~H i.n !"~; '3 >" T < LO LO xT Ul Ui iji Ui UJ UJ UJ UJ III 03 CM i3 K CD CT| 0,1 iTi ■t— i Xi \! rn •t* ■"vf" CO ro -.— < «— = IV- CD xf CM cn UO 'X •<— * 'u ■r-1 !Jl '3 0"! 0"! !-- IV. i n - LO tf ■^r ■^r 1 "\T V Ui ui Ul Ui Ui cr. I ■ : '3 r ; CD C" '3 •3 13 13' CD '3 '3 C ) 3 -*■ - 1 -' ■^ ■^ - L -■* -^ - 1- . - J UJ UJ in IXI UJ IX UJ UJ UJ LJ UJ UJ UJ 'X V {■•- ro Xi ro '3 ro O'i ■«— * r^_ CM ^-H Ui r-- i-X IV Oj '3 CM ro ro ro '3 'X Ul r- T iTl CM ros ro '.jj IjO 03 Ul 03 'T UJ CVJ i3 r\ IX' C.J Ul 'Xi CJ m T i3 'X ^H [\_ IX ^r 10 CM '3 Ul Oj i-X in ^r C'J T-H r- r«- IV r-- r- 1^- ■: if '3 3' >3 IV. in 13 '3 i3 i3 in iS 13 in 90 m CD CD iX CO iM CM '3 CO S3 CO 3 <3 CO CO IS G) CO IJJ LU IJJ LJ UJ IJJ IJJ UJ UJ LU LU LLl UJ OO iTl CTl IT) CTl CM »-H t 'X' \r IJ) in xr CD r- LO UD iTl r ■■- [V| ^i - r- ■>— < CD is CTi ro CO I'O CM iv CTl CD ro m U") ■r— ' uj IV 'X <3 'X LH 'Xi iT;i IV f. en vr CM CM ■ m CD CD -.-H m in CO CD ro 'X CO -ST CO CM •~* ■^r 'XI CO CD CM in r- iji C'J Nd IX' CTi •X in ■sr ro ro CM —i CO en CTl CD IV 'X 03 CD CD 03 00 00 CD 00 r>.- IV IV IV IV '3 CO CD CO IS co CD CO co co CD co CO CO CC o Ll 00 h- _l ro co LU Cfc Q Ul > LU — ta. Ql LU I — 00 UJ ui n n ol =d "• in no •*— « ff •^r in in ^r ■n W-H 'X CM IV CM CD in 00 ro CD ro T ■rl- ro M CM CM ■— i T-H CD CD CTl CTi in m m in in m m in m in 'Xj" sr Ul CTl cn cn iTl CTi CTi CTi en cn iTi cn CO CD iS ^r co iS iS IS iS CD CD CTi CO IV 'X in CO CTi CTl CTl CTi CTl ro CTl CM — CTl CTl iS IS CD d CD CTl iS iS CTl 00 LU Cs Hi Ql LU }f jt ;••:; V- >: v> K }•?' ID CO ID CO CD CO 'S co IS in CO IX <3 ITi CO in GO m ^ ^r 'X 91 en l'- ! _J UJ I.L. d h- _ i-'-i 1 t- * 1- .2. i CO i.O I I I LU i CJ I LU ! I Hi I o I t- I i ! _J i i33 CD E' Eli UJ cd G3 E- LU LU LU UJ UJ UJ UJ LU U 1 UJ LU iH 7 3" IV iT'i ;-rs UD U.l ~.r CM CD Ln ro lh m IV ID 'jD 1. '"; «-H .-H m —i i.'J ,_( G"i i rv CD kO ^H rv f ••— i CO |V ■ D '-ij IV IJJ T— ! 'U CO ro L r 'i co Gi *- LO ro CM ro '". D r.j i ■ ' ro OG ro r- CM iv CM r ~ fM r\_ '"'••J IV •^r \r ro ro CM CM -,— 1 .— i CD i'Ti G"i CTi n'i Wf in m LD L'"i U"i LO in LO LJ": ^.r ^r M" i.'j CD cd Ei CD es CD CD cd SD CO CD CD CD CD UJ LU rv CD iXi E CD CD LU IjJ LU UJ LU ijj r.', 0~i ■»— j 1 1 1 CO ~: iV; QIi U0 ro IT, ro \r M cm fM ■^ uo i,jj r-w '.' LTi LO L.' LD f-— rv IV |V 1 '- iD CD CD CD .21 1 LU UJ LU Ixi LU LU LU O 1 LP i— 1 CD '■• LO CM *— < 1 — ' i IV ro ij] '\j CD ■«—i LO -J 1 ('-._ 0-.i r>- ro Ui 'X' CM .-J 1 CM LTi l£l r-- 'JD in ro _ | *— 1 t r- i-^-i ro LO it. .£_ i CO r-- l£i i o W.i t ro 1 cm oj C-J CM cm CM i M — * 1 en >Ti Ui Ui en iTi en LU 'D LLI CM L'J UJ LU 1' - T T L" i"'-J OJ CD M" © in m 1— « i n CO 0? r- i_y I^.J UJ k r- r>- iD CD CD ; i UJ lTl dj CD CD li.1 li_: '?0 E' U.I U.I LU LU III CD ro LTi |V CD iT'i IT' rn P..I m O l.o en f ■ ! in C^j ro in rv M" f- Ui T-i rn OJ '-i iD iD Ijl CJ iD iD CD CD CD cu UJ cc. CD Ei 'D Cv m ITi r^- EJ CD in CD CD CD —< -^ CM ro ^r LO ID LTi 92 O I OD Li.: El '3D E 1 13 K( 1 i s 1 1 1 LU LU IJJ LU LU i%- f\_ r.j cn ijD \r ■^r IV CC C\i i\_ Li J jj~ r- i l - \l U 1 LD TT "\i tf LO U3 Qj •.— U~i CM !"Ti ! j3 ro E) LU CM ^—1 CT"i CD r- ltj '^ m ro CM CM !~sj CM C3 LU CD LU d Pi on o CD 23 CO LU LU Cii LU I- CD LU H OL Zj ~ZL U. O 1- LU z: lu u UJ CM CD i LU i f\l i ! LU ! LU LU 1 =S Q£ i 1j i cn ! i iQ i O h- CJ LU ro CD <^J El LD ff't LP CD U.l i. ' IS 1X_! Ei LU LU LU in T-H r- ',Li CD r "i i.n iji in i\i j- % i f"! in T °\T ■"5" UJ LU r>- ro '■£■ ro Q CM LU i.u LT P- IV m r-~ r-- IS Ei E CTi ro LU Qi a i a- I Lt I LU S Ei is in 93 in ft _] LU Li. <2 ••-* h- >- CC _J Q -jl o c*. o y- u_ ■-' LU •— I Qi 1— h-o.: LU Lu~ ' " 0£ □ 'j" -£-. U_ U. il. CD ~n (fi 2: UJ D 1- <— en CM in IV en '3 i2' IT) CD LP CM CTi T CD CM IV CM jV ro CC CM IV CM IV t-H >Xi CM CM Ai T-M '2' "S en cn OH CD IV IV UJ Ln \n in m LO LP 'T •';!" 'T xl- ■<-r "vi* vr CTi 1J1 en en 1X1 O, en in CHi 0") 1T1 Ij'l iT't CO "3 '2' CD i2' i2 is 12 CD '2' '2' 12 CSD UJ UJ UJ UJ 'U IJJ LU iS CD IT; I 1 1 12 iS 'S 13 2' IS r.-i S IS IS 1'" 13 '2' 1 1 iS IS S I'D 12 IS CD IS IS 3 '2' 13 Q CD C 1 CD IS Ei '", Ei iS CD GD 3 12 12 3 IS 13D 3 Cj UJ iS CD CD (S •V 13 LU LU UJ UJ '2' CD iS IS CD 12' 12 CD i3 iS CD 2i iS IJJ UJ IJJ LU LU iS 1X1 iS CD ® -. iT~i ■^ __ IS3 CD Ct tS3 Q IS G3 j"3Zi 120 iS © CD LU LU UJ iS 3 CD UJ •3 >3 12 S '3 IJJ GO iS 3 CO 3 CD LD Ijl 3 'XJ im ro ro IS iJ'i CO Ui CD CD 13 13 13 12 13 13 Ln 13 - •— * CD ro CD CTi «--H ; JJ \i LO p. 1 LO LP CD ro *— ' LT.i IV Tf -=— : [\_ fO iT"! \r cn ro ro "-> 0*! '£i V — ' j kD XT 1 1 1 i i • [V fv .' '- ! — '-.i- CD k iv W- IV- p>_ IS- r\- r-- h- C" 1 ! ■CTi CTi ■-T-, iT\ C'' Ul C""' CTi O (V o> LL'T 0".i CO ! Lu :r i (Zi o i uj r s > - ! w .^_ ' tv h- i I— a; lu i uj •— n i Qi iXi ~i i.n CD LO "^J" '\T ro ro C\l cj — i «-! CD CD ID 1X3 '£■ l -±.< CD ID '. Ti ijj I.J 1 'Xi l£< iT. Q i G '! IT"! 'Xi iTi O'i i ii Ijl iXl iT'i CD ij'i cn cn 'X. u i CTi IT| 0"'l L'l IT', CD CO 'Hi CD CD CO CD Q CD (Xj CO i.n UJ cn i cn i 1 1 1 1 i tv h- u 1 CO CD ■*-H cm \l 'D 'J; •■j U."i iji.' \1, i-. i CD ■:T 1 CD CD CD co UJ ITi CD T—> T-H C'j P.i ro ro Lu 1 CD cn CO r-- I.J- LD •^f ro ,-,! iT'i cn o:- 1 CD 0"l Cn iji CTi u'"i ij i iji CTi O'i cn CO co _l CD 1< \* ,,, ~* -., -.;. 1 CD ID CD LO CO ID CD LO CD LO CD UO CD IJi rj 1 CD CD «— i *-! 0-i CJ ro ro \f -J- in LO 'D Qi III IJJ 95 ■X --- I— •x _i &-A ■£; rv UJ o ./- Li- X en 1 ft — .- jL- en " UJ X IX -T o l" LiJ X J? "— ■ UJ n-" I— H III DJ I— 1 3 a; m -r 7" Li. LL X o o i 01 "^* III o I- i_. ..- 1 — 1 ■X "7" en ft .T >— i o LiJ Lt 0- o en 0) i — _Q ra ro iu LU iS iS iS IS 'S iS >S Xi iS iS ii.i X' iS iS X' O 1 IX IJJ IX IX IX IX IX UJ Ijj XI XI UJ LiJ '—1 1 ro CM CTJ Xi Xi X' IV ■^r oil 'X UJ Xi IV _J 1 Ci xr CM 'X 'Tj IV 'Xi CM xi' ■^x ■r-l IV •^-i _J 1 03 m XT IX 1X1 IX XT U3 'S [■-■ |V IXi XJ ' — i : iT| cm X! 03 ■r— I !X> Ui rn rn C'-J !'■■■_ C-J 03 XT. 1 XT •r~) IX IS 'X .C-J CTi IJl CM 03 LO ■r— f xi~ P'"l -..- 'S iXi IV 'X XT rn CM 'X 1 CTi 03 LTJ i >X U3 CO 'X IXi IJ"I IX X' ui in IJl ■^3* "■xi" CM 1 *-* "~ ' ,-H "•""i .1—1 ■'~ l *~ l ,— ' ■r-H ■r-l ;~<. -.--I ""' Ci (S iS S X IXi iXi X' 'S' 'S 'S iS 'S IS Xi iXi Xi X 1 — - IX' iS IS IS IS Xi IS IS liJ 'X' IS iS IS X' 1 LU IX IX LU UJ UJ UJ LU IX IX IX IX UJ ! — ' ; •■;]" •c^- en 'S 03 ""\J" IJ '1 m Xi LO 'XT 'X CM X i CO ■r-H 'Xi CM If) IJl 'X' Ij ! 0"'i CM LP LD ■•— t _J i ' r :T CM CD Ul jXl cr 03 Xi 0"l 03 XT 03 IS *~- '■ ] \y r*_ 03 iJl LTi IV xr Xi in Xi 0-J ro XT LiJ JZ i ■ 'S Xi 'S iS X! { UJ UJ IX UJ LU IX IX X IX X IX LU UJ O i XT CM ■<— 1 X' iS r-- CTl iS ' rn iJl ■*J IV C'i ^.-i j iv X' -^ iXi IXi U3 rn Xi CM -.— ; ■^ 03 '\S* _l 1 ]■ - 'X' 1—1 cm •r— 1 X 1 on X -.-- * iXJ X) CM CM U 1 CM ■.-1 'S 03 C£i ro IS r- ■sr Xi Xi '-H kO h ~ < 1 iv !J3 ro X' 03 i.J3 XT —i 01 [\ "\i" l\l Xi "~ 1 p. iXi Xi IV C~l X 03 03 03 03 IV rv. IV IV X 1—1 i i — iJl Xi Xi iJl iXi |V iTi Ijl Ui Xi X Xi Xi ~ l " IS 'S iS iS . ~*" "'"' 'S 'S IX 1— !■ i IS X' •rf CM t LJ3 '7. 0-i '•J3 'S Ul iS i.n ■x IX 'S iTi 'S 'S IS it: " ■-— i ■*-* CM 0J rn ro u_ iS G"i 03 iv IJ3 LO "Sj" r-3 CVJ •^* iX 'Xi IS IS Xi Xi iXi Xi 'Ji cn en IJl Ui 0-i 133 03 -j _j *-H iS iS >s iS IS X 'S 'S S iS iS IS X f j UJ IX LU 1- X }r ?:■■£ 5^ ?i J* ^ Jv* J<: ?■* >* •>* >^ $z X IS IS 'S Si iS >s iS UJ Xi Xi X' iS iS CD Ij 1 IS LO ■s ID iS in X' IXJ IS IJ3 iS IX o iS iS "-I >-i CM CM ro ro x r \T LO U"i LD IX X LU 96 '3 'ZD •ZD CD UJ UJ LU LU LU CD ro ro Q] CD ro CD CM ■>-H • 'IT'' IX. CD CTi ro Q~l - r kQ '-0 LH in ^i .— , ld ;Jl ro |V IS "vT ro ro CM kD \C >X) U3 l_T| C3 iX; i30 '3 S3 1 IJJ 1 LU 1 UJ- 1 r" "-r ro (f CD m P.! iVl IV l~ CD -t CTi rv 1 i 1 LO LO Ij .' CD r- T-H Li ~i |T"' r.-j r- l?>J — i l_jj ciD cm i ij.] 1 ! Li J 1 i- H IV i Li") r- i U 5 Oj i LO f..l 1 CD CM \ 'A: M i CD CD CS CO ■5 o 5 co co o a . LU r-.j LU 1 ' ! LU LU LU UJ i.i.l f\l *-H ^—4 CD fft m iTi iTi ^rr 00 0"* ro CO iv Oil i~ ro or. Lfi ro ■^— i ro CM ro ro «•* IV kD U' ro LTJ IT: 0-.' ■tf ■^r p. 1 — < 0~i r-- ko Tt CTi iv IT. r- CTi f-- 03 rv co r- CD r- r>- GD CO rs cs CO iS i3Z* IJJ Lj Lu ui UJ 1 ' S3 1 1 'i -> — ■ IX ! i — ; | 03 iv r- 1 S ~- ro i CD wj >3 '-jj en i I'M ro -. — i T\_ ! '■ J I Ul ,_-. ro XT I.JJ i CD en |V LT5 m i CD r_ r~ r- r>- i IV. |v iV f- |V ! I_l O > U_ CE CD LU CD I LU X I UCi- i T. I * O I LU il I > "-' I IjJ I ■~i SI I fit I— I hULU I UJ ■— T. I ui iXj ■! I z: i u_ u_ Q£ I o o s i HO I z: u i a I- a i "II : CO i— i CD " ^ kO 03 OD Oj IV m 01 ro r_ ■^H U" Ui ro i>- cd i^. tj. ij'; Oj IV rv cn ro r O ro ro f' 1 '! ro en G", ij-i iTi CTi ij. Ul UJ r-- ui d i v_i ixi ro Ul -a en cu i—- ' J2 CCi 'X' — ^ CM ^r LD Ui i"-J llj CD 'Ti CO l\ ii*i in M' ro CV ! ij'. Ijl U) Ui Ui Ui O'i « 13 '3 CS3 G O G3 Ei U > S i/J IO IS U i Ui 0" *s G3 G3 I'D © ■53 G3 i. r j S3 !J"J iHi \n S' LTi 97 CO S3 LU •I ■- h- y- •:X _J LX O >- S 'X en LU x lii h- Q O UJ H LU _^ ixi- ii; UJ LX tC UJ 1 1 1 1 1 1 ] 1 'X iX i i __ ; I i -j 1 " 1 X , ■ | j -?■ i ix i IX i UJ 1 IX UJ UJ UJ IjJ UJ Ui IJJ UJ LU ijj ! > — i j CO •*-) X 1 -— • on IX .-< r-._ ro v-H r- IJ .i iX ; 1 ( iX ro ■— IX ro CD t*H ijl LO 'Ji 'X Si 0"i ! . J 1 ro 0"'! iX 'Xi 03 m IX [V i? i 1? J iX " 00 l >-" i !_.-] ■ — '. o~. en co rv ro.' •■— ^ ; ^' X i o I.J-. i £~. 1 Li"! 1 U. iO UJ ^— i in CD \t in 3Z» i."U" iX E3 CD .£- 1 LU LU U_i Ul IJJ !±! IX UJ LU LU LU LU LU 1 — 1 i r-w iXi ro CM ■«— ! r- Li"; CO ^yT Oj m -J j L£i CTi ro in ■^7 cc IX lo L". ro iX \T\ 3; _J ! in CM ,_ CO (S :11 — * CM ro '^t X" ^ *\T —" 1 '•X) Q"i S3 U i i . — T7I- LP T ^- ■-— i 'X - *-■ -xT i 03 IX Oj r- rv f>_ IV i.n '.D LO 1X3 ' O in 01 1 LO >X IX l_T; IX l£ j ri '.D ij~t 'UJ Uj '■£i LO UJ 1 IS iXi 'X IS IS X iS 'X' '3 ■X: IS '3Z' IS —• u_ CD 'X iX iX' Gj iX (S3 iS ;7Z' CD CO »3Zi CO 2; \ l_U U'- Ul IJJ UJ UJ !X UJ LU UJ UJ Lu: UJ O ! r^- IXi 'X r-- IV in 03 i£i UD "^j IV ,— j ^-H 1 — ' 1 IXi 'X rh 'X r- s CM Oi i" 1 - LTi ro © CM 1 | in LO m CM ■<:■■ CD i.n ■*~i CD TT Tj" 'X X 1 en if; ■"•j f>- CO CM ro ro ■=— ' 0Z> ro f -1 'J'. *** ! 1*0 ■a- in in Li~l 'X Ln in '-■D Lfi LO -Tt ro T~ | CO iv IX) in \J ro OJ •-i SJ 'Ti CCi r^- L£l IX G'i iX en on CO IX X'l IT- iJli IJJ 0? CO — 1 CO CD CO 'X CO iX GO iX CD GIi 'X Oj CO CD „ ,_ j ,_ __ ^_ 7r , iT-i _ „_ 1 1 G£ K oj iS IX ■^-1 cm XT IX) iX CM j .D '3Z' IX S) LO ■X IX IS (3 co IX 'S 1 ' •-— ■=— • 0.' OJ r-~i ro u. CO CTi Gj iv lo in ■t ro> OJ -^-j 'X l_!'l L'J S CO CTi •X en Ui en IX 'J- Ui iX CC' CD _J - « _J ,-- 1 CD iX 1X1 iX 'S iS 'X '3D >ZL' CO cs IS ■I IX UJ li Ul i— S ■J-i ^* 5-^ Ik. J.* 5>$ ^ Cfc iX' CO IS CO 'X X IS iX 'iJ i^Ti CO i3D IS IS in l-J \r. S3 in iS X iD in X in iS li" o l£ IS -.-< ■~* CM CM ro ro •Ti" rt LO LO 'X a IX Ui 98 Q l_v_l CD GD G3 El Ei GO GO E 1 o 1 LU LU LU UJ LU LU LU UJ LLl UJ UJ UJ LU 1 — 1 j CO i"V! *— * GO GO ro CM ro LO t-H in '•-I- in i ! ro 0". LO <-- 1 CO ID LTD ID CO ro CT'-I CO en rn Cfl ro Ei Uj •X CO CM :T: IV IV E' in i^ iv Ml ■^-1 LH S3 'Xi ■^r r i CM y-'i LO en "£^ i •^ "yf IV ,— t ■tf CO *-H in •Ti ro IV -. -^ Li" CD in CM GO IV ,j CM en I. f! T' *-H IT. 1,0 S3 1 |V rv IV IV KD CD uo m |J".| m m ■$ vr CM ! CM CM CM CM CM CM cm CM CM CM CM CM CM CM I • LO u. a: LU r-j !JJ _l LL CO ei a CO LU X ci a lu z: .> ■--• LU — X CKI- I- £V LU lu — r ct: co c: 2 U_ U_ lii o a ei en z: lu oh- a M az to Q iD E E Ei S3 Ei E G3 j^ y< 3* j.* 5 1 * K >,'• J*t ?■"■' ?■* J.;i V j.* Ei G3 - U- ■I m y- ft _i ^ ' i •I i n i.u 3 Dc 1- y O i i h! ;" *^3. Ill ■ — i 7t* u. I— H l>' II ! 1; >— i ,' 0£ CD ''J. U_ Lu Ct i i a CD '^ UJ ....' i- - ft i — i ■2" .iL. CO ft •rr o UJ a" 0- CT! CO <1> jO i IX UJ |jj UJ UJ UJ UJ IJJ UJ T-H T-* ID i.n CM U'l LH in ■^r ro .,--1 CD 'X' a CM ro ro ■r-i i "Tf en ro 1 Q CO ■*— s e IX |V CD l"*J XT ■\1" cd tr 'rt T- J cr ro CD 'st - ''T Q i u.i "xT E r>- "\t CD IV "l CM CM ro C'-.l I M "T ro f— i \t CM ■^H iTi ? ■sr t-H Ui CD IV (V IV IV en m ro cn "i E IT-; m iV"i l"i ■"tf ro E m o:i iTi r-- |T"| ■*->. 10 LTJ CD r\- CD IT, ui cTi en iTl j rv LTi ro ■^H 0"i IV IX ro ■^H iTi rv IXi J CD r>- rv i.'~i r-- CD r- in IV in r-- in r-~ IX r- m rv IV '\T IV r>- j E c E E E !jj E E E E E E E E E iTi iTi E E ifi Ijj (M O LU IT LU LU 1 •I ' — ' i_ \/_ Qi 1- QC 1 — Qi ! 1 _i LU — H tn SVQI Z LL U_ 111 '_> I— 1 —I 0" 21 LU O h- O ~- CM CM CM CM cm '"t"' GD i3L' ij-,.i LU LU LU LU LU i i\- t— 1 T-* *— ' IN- iHZi in CM — < CH VD -7—! LO E) CD i7"i "r- 1 rM V T~i |\- Tj" GD U3 IS Gj CO CO cn - l._Ti \T CH CT: — ' m Cij Uj LD ID \j" ro r- j l"J S CO S3 CO co C3 CO CO CO CO III LU LU LU III |V IV -J- i r O iTl ■^ Q CD CD CO i.n pm CM T+ cn r^- ■*-< lh cn ro CH Ln S3 LD ->— * r-- U3 LD M CM r-- 1 '- IV IV 1 - UJ Lu LU OTi LO 1T1 U": CTi IV ^ LD 1X1 CM |V ro UJ IV 1X1 *t X' ui LD CTi cn CM cn S3 CO CO in s s CO LD 101 ro cn CO ID LP ID LU in iS ,T-j -T-. |T-j ,-r-. j— . 'S 'X' iS S3 (S CO 1X1 'S 1X1 iX' CO IS III UJ UJ LU LL 1 UJ UJ LU LU LU LU LU UJ nT U.i CM LO CD IX) 'X. CO iTi i'O i'Tl '^r U' ID UJ t-H cn CM cn 1.-J m ■— t iX' LD IV i.n cn ro iv 1 ro LD ID '.ri '.£1 ID Sj ro '•£1 rv IX) \; I" 1 -.! CO ro IV IS CJ ro r-.i C-i I.D IS LD iVi in Ui ^r CO ro IV -,— i ID IT 1 TT CM cn Pw \r ■-* cn in ^r T-i iTi '■£1 ro ID in LD UI LO LD \f \r ■\" : " M- ro ro m TT *tt ■\T M ~j \f \t -sT "J Tt "$ "^r 'Xi X' 'X' 'S CO ■X 'X ; s CO 'Xi iS 'X' CO .1 u LO 1-H ro X © LO ■sT —J T— J CD \r cn ro O- © ro m N- iXi CD X -J ^r X' U": Oi OJ U3 ro i-H *— i ro l'"~ ro -<-^. 1— H -.-- 1 i— i *— t ro r- -<—* i - xr OJ T-H ;-. ro U3 !X LO t-H r- ro X 'X OJ 0"i "JD ro © r- XT ■^" < CTi U3 XT *-H 'Xi IV "-| OJ X on m ro 123 NT ro f'3 ro ro OJ OJ OJ OJ OJ T~\ T— 1 ■>— i CM OJ OJ OJ OJ OJ OJ OJ OJ OJ OJ OJ OJ OJ CSD © X x X X X' X X X Xi Xi X' X' X © X © © © X X . © Xi < 1 J X i:.,i X Xi Xi X o X © 'X 1 X UJ UJ LU IX IX LU UJ UJ LU IX li ! UJ X <— - : ro LO CD X LO r-~ cn ■r- 1 m LO OJ •*— i y?_ ...J oo l£j OJ rv CTi X © X CTi rv '.13 in - T in © i—< '^i' cn l\l t— ^ iJl t— i X LO U. T-H fO in Xi \t OJ ro X ,_J ro ■»— ' xr OJ 'JD rv ro in ro © iXi 'Xi © ._! X ro LO r-- © cn X © X CTi iS r>_ IXi '-H en '£' I-"' © r- ^ OJ cn l£i OJ cn 1X3 ro XI r>_ LO ro ■»-» 00 U3 n *-H cn r>- XT OJ X : ^r xr xr ~r ro ro ro ro OJ OJ OJ OJ OJ "" H ro ro ro ro ro ro ro ro ro ro ro ro ro X X X X X X X X © © is:. X Lt. a h- '_' X X H OJ -r 'J3 in 01 'Xi X LO X 'JO X © © X o © © © -.-< 1—1 OJ OJ ro ro Ll X en ' II rv 'in m -ff ro OJ *— * X cn X X en cn cn cn cn X en cn cn cn © X ...J _J 1-1 1$ © B © iD © © X © X X X a f 1 LU X r c: >? 3.* j.:; s<> ?'"■ « ?'* »s j.* >z >£ *: N i\ X S3 E) © in X © X © © © © © X m ,t, in © 1/3 X in X in © in © &, Q X © —j i-H OJ OJ ro n V Tj- LO in Xi fX Ik LU 102 ro ro ro ro ro ro ro ro ro ro ro i I I (2 ro ro is ro ro ro lS ro 't' roi Hi LU LU LU LU i_u ijj uj LiJ LU UJ U! LU 1 [■._ rv \T IX! r-j en *~ ! ro CT'i CM M CO m i ro ■rj- \T sf p.! U: in CM r- ro r- roi ro i i. 'i |T| X' |V CO CD ro 1 1 1 ro* i~i iTi CD ro ! ro *~i ijj ro ro i.n CM rn CM ro CM r\i U.l i ro Xf IT- PO CO ro ro. Ul ■--: i •_ 'T -r— lit ! ro ^r CO ro r- ro ro •r-H 'Xi ro UO roi %r i i.n LTj xr T fO ro CM CM '-I t-H CD ro. iTi 1 lil fj"; U" LO LO LO LD LD in m UO ro ■sr i ro ro ro ro ro ro ro ro' ro «s ro ro ro i Li- ce (-: rr ro I u_ ro ro ro o-i LLl x Cfc t— LU _ _ ! Qi!- i hKUJ I UJ f-i H . Cc ro, ro ; ll. u_ E I ^ "m i 2: LU i '-' t S I w a ■! ! LU LU LU _i LL CD (ID CD E CD roi UJ li i LU LU LU ro ro '■J_' t\1 ED ro LO ro pi CD -r-n f \ i r>- t ro ! -J3 ID i.n CD ro co ro T~* ro 'X' -■— • ro ro ■*-* •r-^ 1 \! ro LO ro CO i-r-i '-U ro S3 n- XT ■«— » CD f.T( CD (Jl UJ 00 CG p_ CM '"■J cv CM :.M CM r-j >5J >3j GD •Zi ro ro ro LU ^ ! UJ I UJ i UJ i ! LU I ■-— ' CM ro^ XT ! lV: U.l ro \n i ro i ro i CM T r-- CM ! \] '.O i~o ■r— < j ro Ei p.- LO i i - r-- ro ro ! CM r.i CM C i i ro 1 1 1 i i ! 1 1 1 iTi ro ro ro G3 cd 1 'j 1 1 1 1 — : 1 i ro li.: Gj ro ro ro 1VI ro '--' ro ro ro ro i i LU UJ LU UJ LU ro ro UJ LU UJ UJ UJ Ll' l ! ro -r— ! ro LO U"i ro ^— < p..i LO CM 03 ro en i ! ro ro r-- ro r>- ^ -^-f ro ro .J IV ro ro i i ^— < 'xj" ~. en -.— i ro \r xj- 0"! ^r r*ti .^-i ro i ! *-< ro '.i"i 0"! CM UO ro ro r>- pj IV '£> ■■— ' i 1 00 CD ITi UJ ro ro LI ro GD t~ i .^i 0--J 0J i j \T\ CQ [\_ i.n LO ■*£ ro ro CM i-« ro CM CM ! i ro PTi ro f^3 ro ro o ro i^~: ro l 1 ^; ro 00 1 ■-■ _ — ' — ™ — "-' -"' '-" '-' ro ro i i CO ro ro '30 ro ro ro ro ro ro ro ro ro i ro UJ Q_ ■•—> ro r CO UJ < Qi 0_ uj LU> Sf *-• LU — ' X ac i- t— a: uj 1 j i t — • •c- a' m a X li- u_ ii-. o a -j ij-i 2: ixi a i— \Zi H-H .X ZZ CO (-1 ,'T; UJ UJ Q. a ro <=d- 0) r— -O ro UJ as a U Uj UI LTi in |-J3 IV 00 IV UI CM 'J3 1 '..0 N- IV Lf3 *— i ij"i IV IV U " T-l LD r~ |JD 1 in CD U3 ■^ CM iTi >D ro CO iv ro 01 LO i 10 ro C-l •— < CD 03 IV 'u3 ITS ro CM Ei i.Ti ! 03 03 CO 03 03 |V IV rv IV iv IV IV |J3 1 IT) 03 03 03 i33 03 03 03 133 03 03 03 03 1 ■ ' 1 iTi iTi iTi CTi en iTi CT-, iji CTl Oi iTi 1 CTl CTl Ui CTl i3i Ui ui Ijl 01 01 01 u'i i;n i CD CD CD if J4 CD CO Ei Lf.i CD CD CD Ei E E E CD CD E Ei E Ei UJ UJ UJ CD E SI E CD UJ CD Ei i53 III LU UJ CD CO CD c ■ CD E 'J3 CD CD El © CD CD CD CD El i - ' Ci co E' E lu Ei UJ Ei Ei E IjJ UJ UJ CD CD E E' CD E E' CD O E E CD E E CD E' — CM CD Ei E CD CO iTi 03 IV E' CTi iT) El en Gji \"\ ! \n X CTi CTi Ei E CD E E' CD E CD E CV1 V LD 104 UJ E' E E' LO UJ ~': IJJ cs E) CD i_D 12J Ei o IjJ *jj l-J 1-i.' i^.. iTi © 'ID iSD i^Z- O »s;i S3 GD Gj CD CD T-H •■—1 -f— 1 T— 1 i— i I—! -*— i ■i-H '•J3 E LO CD 'j3 — < CM CM 13 ro CM — E Ui 03 Ei iTi U) 03 03 CD E E CD V" >■> >£ V >* ?': J-J >? ?■? 3>? iV' E El CD CD CD E CO ITi i^ i in E LO E LP CD ITi E LO CD LU ■-■■X CtLU LU> _^- w LlI T Qil- t- ii: uj UJ ca m T u_ u. o o _J (A Jr UJ o F P 1 — 1 ■X X 01 Q a 1 — • LJ UJ Ct' D_ -Q ^f O) r— JD 03 u UJ : CM l CM UJ 00 LO in iS IV CD S3 G? iS U'l iv r>- |V IV LO Ul Ul IT, 01 cn fS cn UJ 1 1 1 LU LU UJ UJ UJ CO 'S nT CM ■^r IT. IT, Ul V-" \T IV — i LI cn C\j — J lO 'T cn CO CM ■Si 01 rv Ul CM IT: 'n T-H rv -T-t- t-H 00 \r -r-H cn m CD 01 rv rv IV '£> LO Ul LO LD L0 L0 01 °J ijs iTi IT, IT, 01 CD S CD CD s CD iS C: CD CD iS iS CD CD IS CD CD CD IS IS s iS IS 'S CD 'S .£i CVJ LO CD ro in L0 LO in ro ^ L0 ■■— i LO CD Tf 01 ro IV -r-* Ul 01 ro I—* CD IV! 01 CT', iTi rv rv i. n i.n LO xr ■tf L0 LO Ul vr vr \i vr i- \T v| vr vr -nT cn 01 01 LTi 0"i iji Ul 01 01 cn 01 01 01 cn 01 cn 01 cn cn 01 01 iTi 01 'X, l i > cn cn 01 cn 01 01 01 Ul 01 U'l 01 iji l~'l '01 s iS s iS IS CD CD IS CD CD 'S IS iS CtJ CD 'S iS CD iS CD S s IS UJ '-J (D o 'S s IS S CD iS 'S *s CD -* CD Ul Qi rv LO m ^r 1*0 S Ui 01 cn Ul 01 Ui en l?J Ul CM ro >s iS iS s LT) iS IS iS in iS iS IS LTi CD S CO LT) IS CD s in IS s — ~ CM iM ro ro ^r T]- L0 ro ijj 105 a. a: y- ft •T ft c- z CL ■•j: o ll ■. r inh h- r; _j ■ 1 :p L ■ji LU •. iK QC ■X ft LU Ul Qi IJJ I — I CJ UJ CL. O •=3- CL> TO U- Qi o 3 UJ (- ft ft -a: CO CD iS •s iS 'S iS IS IS 13 '3 13 iS 'X 1 S3 iS iS iS IS iS iS s 13 <3 13 i3 "7* O LU LU IU LU LU IU 1U UJ LU LU UJ LU LU 1 — 1 03 iv CM CD ■<— 1 ^r IV IV ro Ul LO LD ro 1 (V CM "\" 'S 03 LO CM ■^r (3 'Xi CM ^2" '3 _J iv 1.T1 03 Ul 03 Ul r- -<— < CM CO 1—1 Ul ro '.—■* Ul t CD 1— 1 CO CM ■■—I 1T1 LTJ Ui ro ■^ LD T" •JO r-- |V 03 03 03 iS IV IV LD LD ui \T Tf ro CM ■•— * iS Ul 03 IV <£< LTi '"3" ro CM C' Ul 1T1 1X1 Ui Ul CD 133 03 03 03 03 ij3 03 CM CD CD OJ CO 03 03 03 03 Oj 03 U3 CD UJ '3 CO 'S is iS iS iS iS 'S CD O 13 IS CD 'S iS 'S CD iS 'S iS IS i3 '3 iS IS IS IS IS iS 'S IS CD CD 13 '3 Q UJ IU IU UJ IU UJ IU LU LU LU UJ UJ LU f— » ro 03 IS Ul LP LD ro - r-- xr U< CTi IT! CTi Ul en cn Ul cn cn Ijl G"i Ul tn U'l CTi cn Ui Ul :Tl Ui Ui Ul Ul cn 0-1 Ul UJ 1 El iS IS ■s iS iS iS is' iS C3 13 3 13 LL IS iS 'S IS iS S ts CD iS Ei CD CD 13 '33 'S iS CD S3 s GO 'S •s iS 13 CO 131 ~n IU IU IU LU IU UJ IU LU UJ UJ IU UJ UJ i_i ro r-- I' -1 ! CM ~H I-.-I LD iS LD l?v cn r- LO »H ■«— 1 '3 Ul p- tf IS tn £ 3*£ >»> J£ 2£ K 55 Ji v> >4 J4 Qi CO CO iS 3 iS CO CO O " en Qi X) 5 u_u-cy O i-J i CO 31 LU Oh 1-3 •—■13: Wfll : — i O LU Oi. Q_ ■o *d" O) r— -Q fO -J l~'J LU ^r i_j u LU H- X ro |\- UJ en ro CM r-j en \r ro LU N- LO '"■J CD CD CM ro U-i FO _ v \t 'S ro fO -— IV. CD Si |S- l.'J CM r-- CJ r^ 01 'X S Xf fv. -r~ XT 'X Ul r-- in xj* CM T— 1 cn CTi IX rr ro ■— cn IV LTi LO LO LO in ^r \j" Tf ^r ■\j" ^r ro ro (D CD CO CD CD CD CO CD CD ijj CO CD UL' Xi cn cn Xi cn cn cn cn cn Xi cn IXi U-i X'' cn cn cn cn cn IT, cn cn cn cn Ul Xi IS IX CD LO CD UJ — . cm cm ro ro l\J — O iTi CO cn cn cn 'X is IS •s >t K 3>.t >* }-* ;-r- }.» >? ?.* 2*t ^* a G> CD iS *D iS iS IS s - "": •w- UJ *~ = _L 0£ 1- 1. X X UJ .. QC CD ■:i T. Li_ LU sa i_i L ' CO "' LU O 1- ft CO ft •X 1 — 1 (_) i.1.1 a D_ 0J •Sl- tU I — JD 13 LT3 LU _! r '1 UJ LU CM CM Hi h- x a 0£ LU CD GZi CD CD ITI CD '33 CD CD CD El CD CD CD CD Q CD 'X' CD CD CD iS l.i J UJ UJ Lu LU LU UJ UJ UJ UJ UJ iTt ro I.O \T LO U3 ■r-4 i.n *-H X; CM I'i LD 03 "m" 1 1 1 ^H CM CD IS UJ uo 133 CD CD ■*- » LO CM CD CTi ■»— t ro iv <:r 03 ro 'X! LO Ll") 0\l Xi CD *— * CD 10 en X C\l 1 1 1 ^J" 'Si in *-H Xi ■^~! Ul U3 n 0. > 03 |V UO CM 1 '■■ I 1 ■■■- U"l OJ 01 CM OJ ■t— i — - *-H — 1 *— ' I3Z' |T"I I.l3 X LO UQ 'XI 'X' LD '.£1 'X X' LO CD CD CD 1 El ■•a,' '33' CD IXi 'i.l CD CD UJ ■X< 1 UJ LU LU XI UJ UJ LU LU UJ UJ f 03 X' ro CD Xt Ml IV f..l ■— - LO 1 I- ■t r-- Xi E ; CD Xi 03 LO IV 1 ~~- L3 00 ■^H ■«^ 'JO r-- 03 Ui Ul 1 ro CD IV LO OJ 0"1 kO ro 13 IV ! V Tr 10 ro ro CM iM ■M OJ *"■ Ijl IT, iji IT: r>- 03 1 cn ij'i 1X1 i Cfi C3 03 Ui III Ui 03 Ui 03 03 CD C CD IS IS CD CD Ei Cfi CD Ui IS Ui CD Xi Ui OJ 'X' CD — 1 UO T ro Ul Ui Ui CD !/"> OJ OJ E' CD CD in is iS U0 X' Jt iS iS iS in r<3 s 53 CD iX' IS Xi 1V1 1X1 Xi IS iXi 'X' "tf — iX 03 r^- IX IX 03 1''- -— ? 03 — < 'i-i |v I-- IT, Ui 03 IT, iS O Ei Ei iiiJ CD 1X1 is in ei in in ■X CD 108 CO _J UJ a ii. « w ;~ -j a _i i__ UJ LU * Di lii X :> UJ tt: I- i h- £k IjJ I UJ — H I CL CO •! I LLlLQi I CQj I cn i Z UJ I ai-fl i i— 'Cc z: i co a -x i u LU 4- r 10 CD CD CD CD CD CD CD CD CD L^' CD CD CD 1^1 CD CD LU LU LU LU LU LU UJ LU to |V 03 \\- 1T1 xr ro CD ID \- CVJ tn xr CD en CD CD CD r--i CM IV >D CD cn in ^T CVi GO CM 1 r. r- •j-i LlJ CD in 0". sT CD CM UD IV in CM 1T1 I s - xr CM en -3- \i T-t" Wi |V) ro ro CM in LO m m LO in LO LO CD CD CD CD CD CD CD CD TO LO CD CM Iji CD CD to CM in LO CD CD CD LU XT CD S3 CD CD CD CD CD CD CD CD QJ 'X' CD CD CD CD CD CD CD CD ID ID CD 'X X' 'X 1 LU Ml UJ |jj LJ LU UJ LU LU LU UJ UJ III lj ! ^ iv ro ••£1 CM 'Xi ro CD iX' iT'i CM ro *-H CD cn I—* CD IV CD CD if: IV CM |N-| rv U"J ■— ' ID iii CD U") *-* \j xr CM CD •»— 1 %— : CD m «-H ID i—» ID *— ' Li cn ro kO (Sj ro CD xr ro T— 1 CD CD fV in ro CVJ ID IJ) rv LO in LO in in TT xr "=j \r ■\f *\T ro ro CD 'X' CD CO or. CD CD CO CD 00 an CD m cn cn cn cn cn 0" 1X1 cn U U ! cn cn 0) CD CD CD CD CD CD CD CD CD CD X' CD *X' ID CD CD CD iD CD 1X1 CD cd T' 0") U'l cn IT) CTl lT"! cn cn iT) IT'i IT) cn •T< cn cn ij) cn Xi ID CD 1^-1 ■<— i ^-1 CM p-.i ro ro rs- ID in ■^ ro i"J — ' Q X; 'Ti 1T1 ij) Oi IT, cn cn cn cn CO iD 3*2 jg K 3^ >? ?•? ;■? ?■? K 3-i 3*S }■■: CD 'D iS 'D iD CD iD CD tX IX ID l'. ; iD CD m 'X in IX. m CD LI CD m CD in iD CD CD — — CM CM ro ro XT T+ LO LTi '•/J 109 en UJ t — i Li- ar --X HP LU _> ■*-' 2_ UJ in U_ U. o; O O IX) g£ « — i - |\- in CM in LD CM cn CM CD in _i E Xt r-4 -^■-i in ro LO LO T— < ro LO in *fr _j tf LH CD CO cn T CM •\T co CTi 0.1 a\ ei » — i w cn CD cn CO IV LO CM cn -■J E 'Sj' m 5^ IV ro '2!' in ■«— i r~ ro cn "tf CD LO ■r-H m CM CD CO LP ro CD CO in ro •— i CO 1 J.~l m CD CO cn CO CO iXi '3D N- IV IV IV UJ LO in cm ro ro ro ro ro ro ro ro ro ro ro ro ro cm ■ CD LU or CD CD CD CD — CO CD O CD CD CD CD CD Ei E CD E CD E ra Ei CD Ei CD Ei CD CD E CD CD Ej E CD E E E Ml LU UJ UJ LU LU LU LU LU LU HI LO LO eg CM ST CM r^ *— < LO CM CD in \r xT T ro ■»— i 'Xi cn IV T— < ij'i cr, cm T— 1 LO IV ^ LO Tj- CD CO CM CM 'XI en i— 1 ro LO 'X! r\- IV |V r- CO in CM CD IV "tf r- 1 on LO CM cn CM cm iM CM *-* --• *-* CD CD CD iji IV IV r^- r-~ IV r-- IV IV iv T- LD cn 0"i CTi iTi cn Ui cn cn cn U'l ij E CD CD E E E E E E CD E CD CD CD X J4 J4 X »•£ X X X J4 Jt CD CO E E CD E E CO E E E LP CD in E in E in E m — < •rt CM CM fO 'J -vf 110 E in LU CO ro CO ro CTi LO E E E E 'lJ- E Ei E e E E E CO E E E E E E E E CD E E E E III LU LU UJ UJ LlJ UJ UJ LU LU UJ UJ III '-T E — H r^ ro E ro f~ i CM CM CD CM f- CO U1 Ul CO IV LO *-' LO cn E cn LO E cn E E Ui rs- M" E' XT IV E CD i-J CO cn ro LO CO ^H M" r- Uc — < tr LO 03 cn ■"H ■•— < CD Ol cn CD iv '•£' LO m Tf ro Ov| ro ro I'M CM irvj CM CM CM CM CM CM CM CM cn cn cn CO cn cn cn cn cn Ui cn CTi cn cn CTi cn en cn cn CTi cn cn cn >Ti cn cn CD E ~. CM ^T LO cn iM LO E LO E LO E E E E E E E -^-4 *— i CM CM ro ro E cn IJJ IV LO in ^sr ro r.j i-i E U) 03 CO iT) cn ■Ti cn cn m cn cn cn cn 03 0j E J.;- J? E E IT) E LO in a UJ i — • Li. ~ ■>- 'i _-2* N - y LU 1 — ' ,T" Qi H h- tK 111 LU — . (V 0j - z: U_ U_ Ql Q O * U'' 2Z LU O H M *~ • d r CO ct XT iTi CCi ro LU LO ro © LU CM PO CM \r © is LU © LU 00 PO l£l 0"i PO LU po ro ro lv rv, ro © © LU LU ro Lf) LO ro © LU ro i_u in CD © CD © © ©' CD S3 © © CD 13 13 © © © © © CD CD CD © iD 13 '3 © Q © 13 ©' in LU LU LU LU UJ LU LU UJ UJ ■UJ hi kc LO CD ; w CVi r-- 1A.' on Tt CM C"i ro rv lT.' rv 13 ■KT LO ro ro !.jj CO 'JJ cs in i*- IV kfl iM ifl m r^ ro Uj 'X' ro ■Ti T-< 0-1 f> --« CO ^r CTi ro LO i.n uj LO CD en CD »-* ■»-* ©I CM ^r '£' a- CM S3 ©1 iVl i_j 1H rv t © © © © © 1X3 O in LO © © © © © 1X3 © UJ t'O © no uj r- u, LO iM '■i' ■^-i ,— i rv iji r- cn © in iD i-D © Ui Ui OJ CO © LU a ^> ■*-' UJ CV £ f~ LL' Ixl UJ >-> H 0Z fD ;X U, U_ 52 o a s to -z. IJJ O i— o en o -x LU 111 Q_ a) .a ON LU _1 i_i LU UJ Qi CO CO 1 CD 1 iS iS 1 CD I 'S 1 IS CD 1 CD CD 3 CD 1 ; LU UJ UJ UJ LU Lxi LU 1 UJ 1 UJ UJ 1 UJ LU f UJ h— ; 1.0 CO cn r>- *-i 'X' *— 1 CD LjD cn rH CD ■^r >xi LD ro •X) CM T— ■ r-~ CD ro CO 0. "\T r-. iS IT. •X CD CD ro r^ '3 CD iS LO ^h LT.i :-'-. ro |\, IT' LO IS *X CM 'X CO xr ■q r^ CM 2; 1X1 ro CD ro On Cm cn 1X1 cr, 03 iS cp CO OH IS cx« l : n CO IS IS 'S CD CD CD iS iS CD CD 'S • 3 CD iS iS CD CDi IS S iS s CD CD Ei (Xi CD O CD CD CDi iS cs 13 CD 3 31 3' ~7^_ O LU 1 1! LU LU LU LU LU LU UJ Ul LU UJ IXl 1— 1 LP CO Ul CM ro ro CM •Xi -x CD in tN- CM .J ro CM r. CO in ro IV -X ro CD ro CTi CD ._! in CO T-* r-i m ro -- 03 •<— i CD \r K-. 'j cn ^H * — O: 'X! ^-1 ^X y ro IS 03 in *-< CO LP ~-< r<- ro cn U) CO N CM *-H cn iS r>- in *X ro *-H CD iS rv. CD "X in LO y+ T "tf -x -X ^x *X -x ro ro ro CO a; CD co CD IS CD CD CD IS CO CD CO CO i_j CD IS CO CD CD co CD s CD CD iS S3 _J u_ CD CD CD iS CD CO IS IS CD ED 3 CO (S3 s iS IS iS CD iS CD G2 CD CD 3 CD 2; LU UJ LU UJ UJ LU UJ UJ LU UJ UJ LU UJ TO ro CO 0_i CM CO ~x 1X1 -■— . K- ro r\- CO < — ■ — * Cm Oi — 1 "X in ID ■^-t -\r CDi CD »-< ro • ro ro CD CO ■— 1 'X cn -x CM \r cn l\- CO _) r- r»- 'D in *X *~* CO IXi •~H co 3 ■x r^- 1 — 1 a co CM CD "X 3 in ■»-< r-- CM 03 ro iS 11 en 0D 03 r-- i\_ r\- in 'XI m IT: *x 'X PO LP LO- LQ LO LO LO in in in m in ITi in — ■ CTi ON cn cn cn Xl iX cn cn cn 1X1 cr-. •XI iS O CD iS co CD CD CD iS CD CD iS i!t. a 1- CD iS •■H CM "X 'D cn CM KB s in CD CO ■•X a CD CO CD CD CD iD ■<-< *** CM CM ro ro u_ CD 0*1 0:1 r- '£1 in -x ro Ol •^ CD ix iS CD Ul cn cn 1X1 cn cn ix 0*1 cn cn CD CD _J _l ■— ' 'D iD iS CD CO « *f: 5-'.' i4 *j >e }>• a: ID iD CD Ei CD CD 'S CO iD iD CO •s CD CD in CD m CD in CO LO a in CD in 'D n iD CD 1- » — 1 CM CM ro ro "X ^r m in ■S IX Ix. IXl 112 CO LU O X lK-X O Lu » X Cfl I— h- 3: -J o nr UJ ■. « X f-o;uj uj — n CK CD q. X Lu U_ Qi 0=) CO Dh « CO O T CJ uu CK EL 1> JQ (T3 CD ca cd CD CD iZD CD CD i_il' CD CD LU LU LU LU LU UJ LU UJ UJ LU LU LU UJ 1 — 1 *-i L0 LO 'Xi ro CO IN. CTl ^r UJ CD IV r-- _l ~— 1 CD T \T ro |S- Uj CTl CD ro ro en *— 1 cn 1 w-l *-* *-• CM ro LO IV ti.i Tf CO CM 03 h.i 1 — i en ~i CD CM U 1 iT. ^r ■r-^ on UD IX) US CO V* CD rv. LO CM en "JD ■nT CM CH rv m ro *— < 'J? ro ■*— en U3 ^T CM 13 rv IX) ro ■-—i U'i :Vi ro ro 1-0 CM (M CM CM CM n-H ■r— i 1-H ■r-l CD Xi xr ^ cn Mj ro *— ' fv '£) 'Xl '.X) UJ LO LP LO LO \r vt* \r V m UJ LO LO LTi LP LO LO LTi U""J LH LO Ul CD CO S3 CO CD CD CD UJ CO CM r- 13 m 13 LU 0J ^r en ro CD CD iD UJ r- LO IX) IV 0J ro CD LU ro uo ro r^ ro OJ LU CD C- kD ro cn LO ro CD CD CD LU rv- CD LO ^r kj CO UJ iS cn rv iD ro ro UJ en OJ r*- r^ LO iD CD LU ro 'XI CO CD ID CD ro OJ CD CD iD CD ■~ CM ^r U3 iD CD CD LiJ Q K iD CJi OD r-- 'Xi IT CD 1T1 en CTJ CO 1T1 IJl ro cn 'Xi cn Ei LO CM CM cti cn CD iD iS CD IS '3 CD iD 'S LU Uj '.£■ CO LO CO IS CO CD iS LU cn iS >x> IJO IS rv CVJ 13 CD ID ro CO CO i3 J* >? +i >; 3*2 CD CD 53 CD iD iS CD 'D iS CD U") iD in CD in CD in CD iD fD « « CM •■"•.] ro ro "vT CD CD ITi in 113 Ul ■x UJ •— . T" CK osF UJ l— a: ixj m uj lu <-« T. .;.. i_ Qi CO X i X 2i X X u_ u_ 0£ o o 1- i~ X' UJ uj en X UJ UJ UJ O 1— Qi lii ft «-" X !- 1— -i_ X< iXj UJ Ijj Mi Gj *Xi - E IS Ei Ei E' Ei ID E E Ei c E IS IS Ei E Ei in IS IS E l? : UJ I IS ! Ei I UJ UJ UJ 1 '±' '5 El 1 iS Ei -< ! Ei I ■ I 114 CO LU ■1 — I ■i — i ■■ — i i—i -.—1 ■r-H CD CD X X X CD CD £3 EJ 6D X ll! LU UJ LU UJ LU UJ UJ UJ UJ UJ X i53 X X X CD <3I i © CD CD X X X X X E CD CD © 13 CD X iV X iX X 12 i£: fTi CD CD CD iS X X IT' X IS CD CJ CD CD C' X X iS X X IS CD iSD "3D © GD 'S •XI X iS X IS CD CD iS CD \~\ X' X X X X G3 - 1 " -v S3 "* IS Cl L'J x x X X IS X IS s IS s X IS IS iS IS X ll! LU LU LU UJ LU L'J LU LU LU UJ UJ III X 'S X X s X X X X X iTI X X X X IS X !S X IS X IS X iS X X X X IS X X X X 'S X X X X X X X X 'S X X X X IS X X X X X X X X X X ■rr\ X IS IS iS iS IS IS X S X X X X X X X X X IS X X X iS in X iS X iS IS iS IS X iS s X X X iS X s 'S UJ IS UJ UJ LU UJ iS iS in CD l^_l >s X I 7- ! iS X X iS |T| in X X iS iS iS iS iS 'S iS X iS iS iS iS iS IS IS iS X IS X iS X IS e- CT» i^T' X UJ LU LU LU iri G? o X CD (V~ rri iTi iS © S X '3D i3j X IS CD GD X X CD CD X iS © Gj X X iS X ■~ 'D ro ^ h- ro in ro CO ii; P0 r> (Tl 'S on 'O ■^r CM X iS In ■tf f-i X ui UI iji 0~i en CI, UJ o:i CO —I !S X X iS X X X X X 'S iS (S3 —, ■— . X X X U0 X iS in X X X X c IT > in 115 L'.l U- >- h- d LU LU LU o h- Ct Q; : — 1 ■ — ' '.L }— 21 i.'" 1 Q CO cn ^E t-i ili Q- a. o LD © Si *Sj S3 'X' 'Xi lV: © E' © © S3 CD © S' EI' 'Xs iX ■X' XI X; © © © © © E' IS S3 '31' IX X- O CS3 © © © © © G3 Si Si 'X' 1X1 I^J X: iX 1 Gj (_!_' LXi CM •r— i ■r— 1 s — < ■> — ' •r— . -> — 1 ■J — 1 f— i - I^J © O UJ UJ LU LU UJ UJ UJ !_!_( t.U I.Ll Ui !JJ UJ i— ( CD S3 CD © S3 © ©' © S3 '1^1 IS iX' 'X' ,_J CD CD © CD © © S3 d ©i ijD 1X1 X: iX 1 © © ij>J iD © S3 S3 Si IS tjili CD Xi 'Xi Ei li-l UJ i T~ © IS © © ST' © © S3 E' 'X' iX> 'X' © r j i •*$ © © CD CD © © © Si © 133 'Xi CD 'X' Lf; 1 **•< "-" 1-1 *""■ *•"• 1-1 -"' *"' "" -.--: ■r-i T-t UJ 1 © ,3 Q S3 © S3 © © Si fjj C"' *X- 'X' __i i i— •» j U_ 1 151' © © S3 © © © © GJ GD >x« '.7Z LU UJ LLJ LU UJ UJ LU LU UJ UJ LU UJ LU o © iD S' © © © © © i3 'Xi iX' 'X' 'X' i — i © © © © S3 'S3 © © ©' i,\_i ■"-"" -* ■* — ' *""' '—' •~ ' •"""' "* s-i "^ •^-i T~l IX' © CD © Si S) © S3 © ~ * X" Xi 'Xi lj£ - — o h- CJ © T-) UJ no -q- |N- r*~i CM rn "TO in u:> r^ ■1 Is t ID © *-■% r'i ro ui i\- Ui i-< ■^3" !"•- X' 1.1- © CD 'vQ ^t CM © CD Ti \r rT) T-. Ti lift _l J © ijl iji iji CTi cn CO OJ ~l CD CO rv- r^- f jj >? "V* J4 ^ 3"S >;> ?^ >t 3^ 0£ © © CD CD S3 © © © CD G? Cj © •X' © LP © m © in S3 LD iD L-~J S Lf- ■x; Ili O CD iS *H ^ CM CM [<< {-'! Xi XT in tr- W C: Hi UJ 116 CO _J LLi h- Q LU •-< X L±l Qih- LU i — r-' ii Qj LU LU —■ ] i_ iZ di Qj - ~i £T Ll Lc Qi I : O h- 1- 3 LU LU 01 z: lu LU LU O h- fv 111 G ~ i 0- I CD ! CJ I LU I Qi I IU I H I •! I CK I i ii: i o I Qi I Qi I IU CD LU CD CD CD LU LU LU LU LU LU IU LU LU cn IT, cn cn cn cn U 1 cn cn m U'i cn cn cn cn cn cn cn cn C', lT, cn cn cn cn cn cn tYi rr ;"r- cn m i"^ )~r. cn cn IT, iTi cn CT, in cn cn cn cn <7l cn cn cn cn cn iT, cn CTi Cn cn cn cn cn , ii cn cn cn cn iT- CTi cn cn cn cn cn Xi LU CD CD CD CD S3 CD CD ID — — LU LU LT' U i IT, 1 IT, cn cn i p'p cn iT' cn Ui CD CD CD Xi CD X' CD CD Xi CD Xi CD iX U.l LU 1, 1 1 II LU LU UJ 1. I IU UJ LU LU LU iX CD iTi CD IX ITl CD 'X' ITl Xi CD G> CD X: CD CO iX 'Xi iX ITl Xi CD Xi CD IS CD IX: CD X' CD CD r. 1 X' CD CD CD CD CD 'Xi CD CO IX' 'Xi CD IT: 'X' CD 'X' 13 CD ■CD iX CD X' 'X' CO CD 'X iX' IX CD •X CD CO ix_r CD CD 1^.1 CD CD CO <-i_t IJ.J Xi UJ CD CD u ro Tj" IV. ro A. I ro CO LO UO rv- Xi X' Xi *-H i" 1 .; ro LO r\- ITl *— 1 Tf r-- CD CD CO '■£i "T CVJ X' Qj '■£i M ro tH ui 01 CD cn IT) Ui cn Ui CD CO CO i_0 GO K- !*_ »H X CD Xi X Xi X Xi 63 X- CD Xi X' It >> N 3\? *■* >J -\.<. a 7,-) A<' j.:> ?? j:> CD CD X X Xi ITi CO X CO X IS X CD CD uo CD LO X in X LO CO uo X uo Xi 117 If! LU i— O --=- •>-' UJ ►- I Qi a: I— ixi 1— Oc I_U CD u,i LU — II L ui co ■! s 'X 2: 2: z Lx. Ll IV O O H I-- 1 UJ LU 0".i Z: LU LU LU O h- Qi Q. 1. — "■Ihl- T 03 h-i CD CO 0: u LU Qi Li- CD ld ai 1 — jTi 1S1 CD CD G) CD IS I..U LU LU Ixl LU UJ Ul iS '51' iS r'- CD iS r 1 CD CD iS IS CD IS cd i-Q CD 13 cd a s iS S3 IS iS IT' CD iS CD 13 •23 CD CD iS 1 CD i^D G' r • CD iS co 'S CD G: CD iS i3 'Hi iS ..—1 " ,—! "^* " ,— i " ' r " J iS iS iS iS Ixl UJ LU LU LU Ixl CD iS iS IS IS IS CD CD CD iS 13 S3 IS IS iS CD CD IS n. CD iS IS IS IS iS IS CD CD CD G CD IS CD IS ,"~i CD CD iS ] ^j CD CD CD CD CD CD '■£' cn Csl ro LP cm iS CO u, cn 03 0J tO r-- cn 'jj ■$ CO OH 03 —• iS iS CD CD iS CD iS .CM Ki 118 UJ Ui cn Ul 1T1 Ul Ul cn cn cn iv CD CD iS iS CD G3 G3 CD '23' UJ UJ UJ UJ UJ <- G3 E) iZIZi i2D E G) CD ■S3 © 'T| S IjL i3D \S) ■1—) '"~ ' ■r— 1 -, 1 : 1 III LU LU UJ LU IS' i3 iS CD 13 CD iS IS CD CD IS IS CO s 13 CD IS i5 cd iS CD IS 13 13 CD iS iS i3 '3 CD CD 13 CD 1x1 54 jj 3-* }'■ #. Si 2*£ X ?•« \: a "*; 1T1 CD iS (S3 O CO C: IS CD iS 13 1V1 CD LO CO LP CO Ln C ' in CD in iS in 13 CO o _l LU »— * u_ ~. 5- cc LU •— ' T" IK C£ \- UJ \— QC LU P ! Ui LU — !1_ ?~ '; o: ijj T 1 IT, Ui Ul Ul Ui Ui Ul Ul Ul 1 ; i iT'i Ul Ui Ui Ol Ul U: 1T1 Ui Ui Ui IJ 1 u 1 1 — ■ ! on ui C'i Ul U, Ui Ui Ul Us u 1 IJ 1 Ul U 1 i .£._ f cn en CTi cn Ul Ui Ul Ui Ui IT: Ul cn 1 en L'l Ul cn U 1 Ul Ui ij ': U"l Ul Ui Ul Ui ! 13 iS 3 iS 3 S iS 3 UJ 3 UJ LU I.JJ iVi 13 1V1 •s Si 3 iS 3 3 LU i^Zi Gj £3 G_i :SJ '3 '31' iXJ KD '3.1 Q CD © '3D l^Ti '3 G3 '33 IJJ IJJ Uf LU UJ tJl 1 ? li! UJ G3 Q iS '33' ! 33" UJ '3D iTi 1^' liJ 'A.' SJ '31' '3D '3 ijj :3D 'X? '3j S3 i\"i i3Ti 'll' ("H '3D Ei '3D 1X1 '3D iT - ! '33 © Gj E3 *3Z* Ijj '3D i3 '3D '33 '3D S3 I.U 3 3 3 3 3 31 — i IJJ 10 V IV PO Ci |V •X (T; LU LU > > ■---■ Ui —' nz Qi CfcL'l- LU 1 — C=i" UJ iXi UJ LiJ '- ;■ .' Ql CD •X XJ X 7" ~7~ T" U. LL Ct: □ o i~ J — X LU UJ 01 T- lu uj ijj O !— a; Qi ft ^ X — 1 — _tL 01 « CO 01 'X 1 — 1 CJ UJ a: G_ CT> ur> O) 1 ID fC LiJ Xi Xi U'i CTi iXi a"l a. U'i CTl X'i Xi LU UJ UJ UJ U3 ID en IT| en U'I Xi X'i IT) Xl |J 'i Xi X'i en U'i Xl Xi U'i Xi m en Xi X'i LU UJ LU Ei UJ IS Ei m E i*v"! i ■ ' IS 'S s IS LU LU UJ UJ j53 133 CO (53 '- 'Us '3 '2 © Xi is E3 iS IS S S Ei 'S iS IS E LU | ■X ;.; 3^ J? >i III i s iS IS s 1 iS Ul iS Ul E IS El S E HI E IXi -■ CM 120 LU Ei Ei iS -i — ! i-H T— 1 — -i i— ■ ~" _u ** -*- _,. - 1 - ■- 1 - 1 - 1 - _j - 1 -' _,. i. ™ UJ III [U IJJ UJ UJ UJ LU LU UJ UJ UJ III '53 © i53 •X' iS i53 "53 153 '33 (S3 © •53 '53 >53 S3 >ID S3 (53 i- (53 155 "53 S3 G "5Ii '53 !.': 1 '53 CD i53 <53 < i '33 '53 i53 "Hit (S3 S3 '53 !_.■ CD '53 '53 '53 '33 t53 (53 iX) CD C3 © 153 IU c -. 153 © iT' 'S3 i53 (53 G3 '53 © © © E ! . 1 53 (S3 '53 r. . GJ i3D '53 iji_i i33 CD '53 & 'SI' '3 © 153 '3 GD '5!' 153 (S3 © © © ro '■\i" r^ ro CM ri CO LP Ul K -r-H CM ro IS IV m ^— i TT IV Ei ■tf CM Ei 01 t£i XT r-o T-* U I CO Ui IX Ui CO CCi OJ LLl CD t<- rv ?•; K 25 >S K E E E E iS S Ul E Ul E' in UJ a i X CD L r. |- 1- -ZL i 'S LU Sil _J if: 1 — i LU « ( ) iv Qi LU ■cc £2± a uj UJ> LU .^ •— ' UJ ■I —i j_ LU Z OL\- LU ct t- \y. lu 1 III 1 UJ HI 2- _ CO q: qd 3! ' i •I z r u. u. |V o a h 1 LU LU ilfj 2: LU UJ Dh i r' (V 1-3 ►-I al- ~x_ tO )S cn CO Ui _ Qi h- h- Qi LU — CY LU Lt- ! LU I UJ CD LU I u_ ll u: o o I— I— ~> LU LU CO z: lu iu ijj o l- n: or q to o co co •! u LU lK CL ld S 'S iS Cj ! -*- -^ -*-' -*■ -*■ ■^ "•■ '-' -*- -" — ■ -"-' HI UJ LU LU UJ l_U UJ UJ UJ 1 U UJ III --h IS 'S iS cn cn i '-I CD iV IV i._j Ul 03 i''" : UJ 03 r-. IV |V IV i"-_ IV p'_ IV iTl 1 ! i iT: Ui cn cn i Ui U'l IJ'I IJ 1 1 1 , C'i >T'l 'Ti cn Ui cn 1 >'l cn T'l IT. i P| IJ ' iJi ITl IT, iTi cn cn IT, cn CTi cn 0"'' IT, CTi iTl ITl Ul cn cn cn cn 'J: C"i iTi cn iTl ij'i cn iT. cn cn u i cn Ul cn u , ui J 1 1 G 5 Ij'i Ui iji Ul cn cn cn cn ij i cn iS IS iS iS 'S iS iS CD IS IS rc*i IS >s ■Ti Xf CD •X C' UJ UJ Lu LU LU Ll' lXJ CD 'X) 'X' CD g j^j IX) 'X' 'X CC IT. >X' 'X' l_£_i 'X E5 iX 1 S3 'X 1 0*3 'X [ S3 CD Cv' UJ UJ IU r~i r- ■ iS iX IS r . ! iXi IS 'X 'XJ 1^1 iXi iXi UJ LU UJ UJ 111 r, ; >X> *r< rxi ITj 'X' »X' tX 'Xi 'X' 1X1 il": s lX CD iTl G3 ^ ro t— t ITl 03 CD CD 'jj 0D IS IS iS 'S IS iS iS in is ir.i C-J CM iS iS in is iS o UJ m r- LU LU iS iS LU 122 CO LU T <=- a .• rv cf O U- * J_ CD h- h- z: _i a x>.^ LTi UJ •■ Cr Q£ O LU UJ > > ■-- LU — • X an 0L' h- UJ 1- IV i:i CD LU LU — • ; > Ol CD ~ i 'X U- U_ Ll Cl Q o 1— 1 — ; UJ LU CD 11 LU UJ O l- Q-' OL l-A ►— ! 1 'X ) — 1 — xz cn « en CO •! i— i CJ UJ a: 0- a; J2 UJ LU _J t — i LU t — i LU 3 co iX CD CD iX CD Ei CD Ei IX' 3 CD CO IX iX CO ■X" 'X 1 CD Ei 'X' CD r o UJ UJ UJ UJ UJ LU UJ LU UJ UJ u.; Uj Uj : — i N 'Xi •j cn |T- cn cn cn cn Xi cn cn cn iXi Ul cn *j cn CT", cn cn Or cn IX cn Xi CTi iX; D i 0i en cn cn cn cn Ul cn cn 'X: cn iXl cn u ■ jT'j LU CD iX 'Xi CD CD 'X' CD 3 3 E' 3 E' CD _l U- 1 U_ E' iX CD iX 3 •X Ei E 3 Ei 3 ■j- LU LU LU LU I ii UJ UJ LU LU UJ UJ LU UJ o CD CO CD iVl 'X CD 'X' E' 3 Ei Ei E: 3 i — . CO 'X' 'X iX 3 3 CD Ei CD 3 3 C> CJ _j CO 3 iX' CD 3 3 3 3 CD Ei 3 E' 3 _i 3 iX CD E' 3 ■-' rf — , •^ •" "~ l 1-1 — ' — • — ' —■ ,-< , -" CD CO 1X1 CD 3 'X 3 3 3 3 3 o Lj 3 _* CD ro ifr N. ro CM i"0 CO in m P.- •X ■X 1 CD 'X' ■•— < 01 ro LTi IV cn ■— i -x r- E LU 3 CO CD -x CM 3 CD ■■D \J" ro i—i CT'i CO 3 cn cn ui cn cn UJ CD CO CO CO r-- r-- _J _J ~- 'X' iX' '3 3 3 3 3 3 3 Ei E' 3 ■I UJ Q£ LU K •X >* *■■: *z ?s ?■? £2 « ?■'■■ ?* ?-* •>■* >-Z ?;: CK CD Ei CO >3 CD 'Xi 3 CD 3 3 3 3 E ! iX> m O h" 'X LTi iX m E uo E> LTi E a: (S3 'X' — ' »H 0! CM ro ro T "IT LP U" l,£ ct: Q£ LU 123 UJ IX o u_ CO UJ L'J Or' h- UJ LlJ .._ DC !x u. K O ;u III i ; i y 1- o > 3>* 3 CD 3 Lfi >D in ■>i :k '3 S3 in S3 CD — 124 in UJ |T| T— ijj 13 CD GD E) GD CD t^) G< rs (S3 '.x^ 1 Gi 3 i_i IJJ LU LU UJ LU LU LU LU LU tu LU Ixl UJ GD CD GD 'S3 GD GJ Gj Gt C i Gi CD GD ii"i CD 3 iTi iTl iVj • ~"; i_D iT"- CD (T", l~l Gi G3 f ~i © '3 CD CD © CD ^j 13 13 'Si! QD GIi i ■■. i ! ! GJ CD i" 1 - 1^,1 ;33 F~\ 3 i3 f'.i CD © CD G' iX> G' G* G' G) '^D 3 iD © i_D 'ID GD C3 Sj G' Gi G' G) CD '3 S t 1 =— i i—i '""* T— t 1—1 1—1 ""^ *■*" "^ M "^ T—i 3 3 E 3 •^-i OJ -r i.n iTi CM 1.J3 3 Ul iVi UD 1*7' IX' '3 3 '3 3 rri 1-H ~-< CM OJ ro ro 3 Ij'l 0J rv '-D in •tf ro 01 i— i •3 'Xi CO CD Ui Ui ui Ul '-" iTi U, iTi 0~'i U i 01 Di ■I 1 71 I U_ QC 1 O — i ! UJ u*i 1 z: h- 1 a •! o 1 •*■ r— to to 1 1 a> 1 r - ■ _Q ro I o l 1; 1 Lg LU CO ro CM CO T ■ 1 1 in -r-i ■j i iT'i cn "T — « rt i\_ ul in ■*^t" 'S -*-> ~~T m ro U~! CD cn *-i ^t — '■ *-3 •* CS CD CD CO •-t cn m cm •TM 'X cn CO iji r- CM 'X CD iji r^ ■~t i ■ i CD T !"..". r-- CM CD "TT cn Tl" cn V u i ro cn CD rv in --r !?. ! ■r— iTj IS 'X in ro CM "vt ■r*- xi" \j" ~*r TT ri- ro ro ro ro ro ro CD Cj co on iS 03 al- CD CD CD 03 CD u. s iS Q3 CO CO CO 's CO CO CO iS CO IS iS o CO CO IS IS 13 iS rr- CO CO CO 'S iS CO CO CO iS iS CO CO CO CO CO iS iS UJ IX LU UJ LU UJ IX UJ LJ IX IX IX IX iS ^ -U cn -r-* *-H o CM in m iS LO LO 'X cn 1-1 ■.' '£> IS *-• cn IN cn cn ■•^ C'-J |M CD in !"■-_ CM CM LO •^~ i~'J in C\i ro 'Xi •■."••J *»i ^- cn ^-i CM IAJ CM T-1 cn r^_ "\^ iS 'Xi UD CM CD in ■»-H r-- r*; cn --r CO 'X C'-J r-- — i *-H 'S o CO cn ij'i CD CD CD N r- 'X |\- r- r-- ••■- r^- ! X 'X I.'" 1 'X 'X 'X 'X ijj iji en cn u ij-i ij) Ul cn cn O'i Ul iji 01 CO CO iS CO CO Cj" iS CO CO IS CO CO CO is is co m is in ■"t ix in 125 lo 1X1 CT h- CE a Oi o LL in h- _j m ijj UJ QJ I- LU Ckl y ix o uj cn zh D CC O co en a u UJ Or: ex. u LO CU -Q (0 I- . Q Ciui CD CO S3 CO iS iS iS IS iS CD IS a LU UJ UJ UJ UJ UJ LU UJ UJ UJ LU LU LU 1— H I'O U3 ro in m in r\- r~ 'Xi U3 ro ■.--I |"< ) _] 00 CO in CM OJ in t -h CO ro cn CD ro IS _J CM U"i r- IS ro LO iS \r 0:1 OJ in. ro ij'i 1 — 1 in S3 'X tj OJ t-H CM rO in a-. ro cn in ..I— Ln ro S3 CO 'Xi ^r CM S3 'X' U3 in ro CM CO 03 lo ro -*— < 1T1 IV in CM iS CO LO XT CS3 ro P. 1 CM CM OJ *- * t— < ■-— » ■»-* •H iS IS is r J CM CM CM CM CM 0-J CM CM CM OJ CM CM CM CD i2Z< CO 13 S3 IS S3 S3 CO co iS IS IS UJ M LU _| u ro uj _i Ixl UJ a: f- C-J a ro T- 1 03 in ro CO CD U"ji CM |Vj; i\- 'X' U3 'X : UD in in in in T ■^ -tf ^3" m LTj in IT) U"i in in LO in in LO m LO S3 CO iS IS IS IS S3 S cd CO CO iS IS 'S CO iS CD is 'S iS CO iS S3 G3 iS 'S S3 iS CD G3 S3 CO iS iS S3 IS CO CO •s UJ UJ LU LU Lu UJ UJ LU UJ UJ LU UJ LU \r nT in IS r- 'S <$ ■•— » |N- cn ID 03 CD •~-* in m 1X1 ro r-_ 'X cn CM ro |T| Ui 1X1 1X1 OJ 03 1T1 ro 1J1 CD ui ro 03 LO CM *-< ro iS \T n- cn 03 Xi CM K cn S3 cn 'X' IN. V CD ! X CM 03 ■tf 'S LO CO 'X IV! m M ■^ CO 92 IN- in ■sT ro T— 1 CO 02 fs. in in in LO ■j ■■? "NT \r ■tf M" \r ro ro ro CD CD CD CD CD LU UJ 03 03 03 CD iS 03 *- CO CO iS iS CD 'S CD •r-l 0J ■^r VD cn OJ ID CO m go i.rt 'S iS S3 IS S3 IS IS — « ♦- 1 OJ CM ro ro iS 1X1 CD (V Xi in tr ro OJ ■•— » S3 cn 03 S3 03 1X1 cn cn cn iji CTi cn iT- cn 03 CD CO iS iS iS s CD CO >': jj \< >* >s ^;: K a J? ?i CD iS IS iS S3 S3 S3 S3 S3 cn CD Lf! S3 in CD U3 IS LOi IS LO ro xj- co 126 1 1 1 1 1 i 1 1 I 1 cm '"\! CM CM CM CM CM 00 CM CM CM '"M 1 1 E i . i '— i E E CD E iS IS 'S E E E 1 2T 1 1 1 i I 1 1 l 1 1 ! 1 1 1 1 '_! ! UJ UJ UJ UJ LU UJ UJ UJ UJ UJ UJ UJ UJ i ' — ' 1 r- ro IV CO CO ro ro '£ T i-H E CO CM 1 .A 1 CM U"i xr CM r- CM Li 1 IT) LO T CM cn 'n ! _J i ro *T LTi l£! UD rv •v. r- |v [V r- ijj U" i i ' — ' 1 in UD CD CD El UJ 'jj E CO E 1. Q UD G ; i ^_ ; '£' *-T ro ■«— i — . s CD T— 1 ~H ro T UD (Tl I i CO s CM •^j- UJ iTl Ci pM t UJ CO E CM i CS3 i On! CM — f CS5 IT'. OD OD r*_ '0 \n \!" \f ro i CM 1 OD UJ CO CD |v r-- iv V T- iv IV |V IV : cm i ■ ■ ■ ■ co _i UJ ►—I u_ UJ CK UJ U- Ch o rj UJ CO Zlh- O 'T. « >- h- Z CO CO 'X U UJ or Q- T3
  • T. in "J f~i ■"3" CM CM Li ' |T| \J- rv rv — h r\- OD 'vT 'JO -rr E "\f IV E \n Ui ro IT) ro en ro \T ro E T]" IJJ IV in 1 in r- 'Ti E CM ro 5T LO in LO LP in co U3 •sr CM ~ cn IV to ro ~-< 0"i |V 1 ! 1 ; .£i Tf 0J UJ E in ro —i iT, IV <\T CM Ei ~-- '-i — i -^ E E E Ei Ul rr* Cn iT. CTi ro ro ro ro ro ro (0 ro CM CM CM CM CM E> CD E E E E E E E E E E E S UJ E Ei E E E E E E ij^i E ITl 'S E 'S CD E E E S E E 'S E E LU LU LU UJ UJ LU LU LU U UJ UJ LU l'l 00 iTi KD 1 1 1 iTl in IV CO IV L"i XT — -. — H •— • LO CM iS 00 LO 0"i Ui ^J CM CM ro m IV E- *-H in U) in r-l OD I.O ~r CM 'S GO ^r ro i—l in lj-l i—« 1—* 03 Tj" CO i~ E IV *— < CD in — i r-- ^r 'S in ~H i£i Ci r^- ■r— '£i ro i—i a. kD t CM Ui r-- ~t CM Ui r-_ m [_r) LTJ T|" T t ^r -^r ro ro 10 ro OJ r-.i UD Uj UD '£! '■£i ! j_i '£i '£i '£i '.£ '■£ '£i '£| IS Ei E E S E E 'S E E iS iS S E E i- CM M- '■£i Ui PM |£' S3 Li" iVi in !^! i^-i E E E |T"| E i—i 1— CM CM ro ro '32' Ui CO rv '£' L£i ■rj- ro C-i i— i E IT"! m E U'l cn |T- U'l en Ui Ul Ui Ui Ui OD OD S3 ijD O K) IS M K j.? >.' JJ 3^ Jv* >•? >.? CO iS E E 'S © E E E m 'S !£■ E tn s LO iS L _ , TJ- E E in 127 CD _J UJ •I a CO h- crt « LU _-'" LU i — i IX h- : i Ul ;> 'X ■ i LL Qi LU CD 7" 1- o a fi — i- Z co oi CD ( — _Q ro LU -r i LU ro CO co IV ro c\j rv '£1 GO I UJ -$ I LU CM ^r |v LO LO cm El CM 'Xi cn rv co ld Ei I LU CM OIi E> ro CO go CM CD I LU LO CD LO i£i r\- co i LU CM rv ro i UJ CM CM iv cm ro CO I LU cn CM LO CO 0-' ft; CM UJ iv ro 03 ro CM CM I LU S3 •3D <— • CM \r CD iTi CM GO GD GO S3 IS GO GD *-H S3 (Tl 03 h- CD in ^ ro CO cn iT. CTi cn Hi u. 0"\ ID GD LT) GO — r>j cm ro CM <-" GO CTi CTi CTi iXi CD GO 03 S3 GO 'S CO ea e> rs — • — CM - 1 " LO m CM CO I LU xr CM r~ s j? >^' j; ;; J4 J-: : ' Jv>- 5i v> (0 iS E' E' Ei E' iS 'S GD Ei iS 53 E E in S LO iS LTJ iS in iS in iS in E 128 CM CM CM I LU 03 03 TO CD CM S3 CM CM LU CJ CO CO CM «-" ro ro ro i LU IV LO ro ro CM I UJ iv LO IT-, S3 ro CO CM ro S3 I LU IV IV cn CO ro S3 I LU Ijj ID ro UJ ro cm cs UJ UJ ro CM •. — i l-o r>- |V U". I'O 03 LO ro S3 S3 S3 S3 '3D S3 S3 cn _j UJ o LU IM CD LU 'V S3 S3 S3 S3 S3 S3 S3 133 E3 ■ S3 CSi '53 S3 S3 CD S3 S3 CS CD S3 S3 S3 S3 '33 CD in UJ UJ UJ UJ LU UJ LU UJ LU UJ UJ III in S3 G*l CJ LO CM IV LO CM i—i LO 133 T-l IV en m 130 !J3 ■*—< CM i— i i33 ro CO cn T-H i.n CO en tt CM ro U3 CJ 133 CM CO ro Tl- IT- LP i—* 00 LO CM cn IV LO ro ■»-l CD 1 , i T-* CO m ■•-n 03 If) ••— 1 03 LO CM cn l.£l CJ en r^ >Q LCi ro CM »— < iji 03 |V m -a- ro Lf; LO LO in LO LO in M" xr c ■fl- -<7J- i i33 S3 Ck a Li- en I- _i =3 CO LU O. UJ LU LU U- O 3D LU CO 2 F o a: o «l-2 cn co a u UJ cir D_ O) LU CJ LU Q 'I 23 Oi 3D CO O F u — * '3D 03 CO to cn •-j- 03 T-H m - [V CM 03 ro i30 CM r- CM CD «-i LP 'D T ro *-* 03 CO ro •— » 03 ID ro -■-< UJ l.n ro ro ro CM CM CJ CJ ■~H »"H -~-* -~H S3 C3 3 T sr ■7 kr ^r ^T ^ m* \r T3 ■ST J ^ '33' i33 '33' i33 E) 3' i33 en S3 S3 '33' S3 13 CD S3 »H CJ ■^r '•J3 cn CJ UJ '3 iVi S3 S3 CD S3 •3 S3 T— 1 i-H CM S3 cn 03 iv CD U3 ■sr ro CJ ■-— i S3 en cn iTi Ui cn cn cn cn Cm — i i3 CD S3 CD CS 3 S3 'D CD ro cn CD F i„ 3 S3 3 LP S3 LP 129 LU U h- _l en LU Ct U.I LU Cfci UJ i n O X ! tn Q O '-Li 'X' iSj CO CD CD CD CO Gj CD CO CO CO CO CO UJ UJ UJ UJ UJ LlI UJ lii ill UJ UJ UJ LU rr «— 1 ro CH OJ "vt »-» ro i. "i %t LO ro CO r- ^r ^r ro CO - LTi TT ro CM *- - -r-H •»— i cm OJ ro CO CM \r l£i CD Cj cm XT UD UJ co CM T i . • CH 03 ',0 '■J-> in \!" ro Oi CM ^-H CO CTi 03 'ID CO IJJ Oj CO 03 CO UJ CO on CD CD © CO i3D co CD CD iS CD CO CO CO CD HI UJ LU LU U_l in r-- IV v ro CM r-i in en ■■— • © ^r 'Xi h- iXi t-« 10 ■»— * KD " ITl CD l w O r*- Pv- ■<$ OJ (Vj rn X< CD CD CD IV r\- OJ OJ CM CM CM CD LU IV CM ro LO OJ U! UJ UJ 111 ■^T •r— i CT; Cj'i CM -T IJJ OJ M i i -. ! Lf3 Hi T CO xr ^r M Oi i.n t CM i^ii Xi Xi UJ X OJ 0-: OJ CM ■sr CD CD CD IS —i **i r- X' in rr ro CM CTi en iTi en en CT. >? »« 3t n J-i 5^ ■>■"■ i? iD CD CD iD CD CD CD CO ID in CD LO iD in CD in CD in CD «-• -^ CM ro ro 130 to (=1 Ge- es LL _l ID CO UJ Oi CS UJ > Lil h- UJ UL CD LU U-i UJ UJ i.O 2: !— o a: c — ' i- x IP CD -1. u UJ 0_ CO a> -O. I o ] >— i cm l CM l\J , LLi O ■I LL CJ UJ UJ Oi UJ f'-i ro ro ro IS '23 is S3 UJ LU UJ LU Si CO r- |\- IV. i'O "\T ■-H T- ; CM CD T f'7 ro LP S3 ro s£i CTi ro ■r— 1 CO ■■— < r-w (V CD CD m -sT ^ ^f "J" Si S3 S3 ro ro ro P~l ro m ro ro 13 S S3 CO Cp CJ IS 1.1 1 UJ UJ UJ ll; UJ UJ UJ [i-i 130 ■st ■s~' •^ 'O CD i5j SI rv u . cm G? CJ S3 Vj (2 GC CO in T ." L^; CM .r-: r-. 'X' i"VJ ~Ti •"■, 0"' ro U'. CS "Ti- Si |-r; Cw -n en M ro ro CD „-i- CTi LO S3 lj_' -r TT ro P0 f '■ ! CM CM -r-1 T-H ■vt "\T \f tt ■•J 1 UJ i LU LU i LU 1 LU 1 UJ t LU 1 UJ i U.I i UJ 1 Lsi 1 UJ 1 LU UJ S3 S3 1.0 m l\_ «— ■ CM ijj -r— 1 \i '71 P-- f"- Tj" i-i en en S3 "\!" cn Co t^J CD i.M £3 CD CD rv CO •-— < CD •— i iv in *d ■st CD Ul CM CD t-H r- ijj ro ro ro [n CD CM p^_ IO 00 'T-" o lp ■*— i r-_ ro cn ir. •^ 03 \j *— 1 CJ tx 1 Gj LT3 |V, r, i CD LO ro ■< — CO CO Ln LT.l -TT T *yf xr I - ** - . ro ro rO CM CM C-J CJ CM CM CM CM CM CM CM CJ CM CJ CM CJ S3 Si S3 CD IT; S' 'S3 Si Si IS S' S< Si S GO S3 Si S3 i\j LU UJ LU LU UJ LU LU LJ UJ UJ IX IX UJ ro ui Ul CM CJ LO ITI CD ~* CiD ro r . UJ LP LO -T CD cn l\_ LO ijl '■D ^\1" CD UJ 1 1 1 ^r CD cn iv CM LO CD ■"3" ^— ; i. '■■ cn t-H CJ T— t UJ CD T— ' r- LO CO Si |\- 'Xi UJ ^J- CJ r- ^H CO CM IV ro 0"! CO CM Ul UJ ■\T C-J *— • PO ■7J- CO r-- cn Si CM Tt uo [■-- cn — 1 \T ro 0-1 -— iS) si Ul CD r- CD LO O" ,, t cn L'l 'T- cn CD ijj CD UJ Oj CD CD UJ ! Si ! (S3 1 (S cn cn 'X CO Si rC ui SJ 'XJ iji iio cn S3 cn ro cn CM Si CJ 'T' 1 C'-..' iTi ro cn Si ro 0-| Si S) Si s< S iS S 1 Si 'S iS S3 IS ! J-': 1 Si 1 CO LTj s s G in 5-2 "S3 iS ><> S3 in IS s LP 3-2 S |X"j iS IP 2$ IS S3 IS LP iS 1 Si Si C'J CJ ro r< - i ■\T •\T in LO l.n 131 _J UJ h- H LU -_ tie CO w J TO CD co ro ro CO CO CM CM $ m en J U i S3 UJ |S- CM vH LI T rv w CM D LU ro i. ^H CO LO LO LQ CQ cm rv ^r CM 10 "T U3 iv U — i ro Ul GO !■. i ro j IV xr -■— < CO LP CM S3 r- "fr *-l CTi IJD ^ 00 CO IJLI Pv. IV IV r- LO U3 U3 IT' in J CM CM CM CM CM CM CM CM CM CM cm rj D S3 S3 S3 S3 ■33 S3 S3 S3 S3 S3 S3 S3 I LU i LO in IV LT3 E) LU ro cm CM CM LO I UJ LQ m CM ui i LU LC LT3 T LU S3 in i LU ro ro "V" CO iTi IV I UJ en LP \T s I LU G3 LT3 LO ro UJ I.Jj" CM CO ro •s CM 3 iS S3 S3 -r CM LO S3 iS 'S s S3 S3 'S 3-? JC J-S J4 }■£ S3 iS S3 S3 s IS Ul iS in iS S3 S3 S3 S3 IT3 S3 S3 »h — 132 i LU r^. LT3 4 LO S3 S3 in m in LU CM LO in CTi ro iS S3 S3 s S3 S3 S3 S3 S3 IS S3 >S UJ LU UJ LU LU LU UJ UJ UJ UJ LU UJ LU ro T LO CO OJ -vT LO CM iTi 0"'i CM LTi UJ r- ro LO ^ S3 LO ■H-i rv ■"T ro 'Xi CM ro ; "*T LO CO ro cn CO 00 ■*— i LO ro ro in \1 S3 CO LU »-i LT3 CM ■>— < ro in CM S3 S3 GO CM LO ITi T CO ro CD ro iS -rf S3 .10 iS LO •Ti ro CO CM r- *— * 'X: S3 in S3 M- S3 Ul UJ CO rv r-- LD U3 m in xi" -T ro LO in LO in LTl LO in Ul in in in LO m ID S3 S i IV! S3 ^H T-! CM CM ro ro ?- i.n LH -r ro OJ i-H iVl Ti CO ■Tl Ti T. Ti CD 0"l 0"! CO CO on S3 iS IS CM xr tj- \r ^r \r \t ^r \i -\T xr ■\t ^r \r C£ CD cd CD CD CD ^ cd CD CD in Lu LU LU LU LU UJ LU LU LU UJ UJ Ll! ■^-» 03 rs- X) ro in *-H ld L'J CTi cn '-i.i (S3 m CD ■r-t CD CM CD Li~' "TT CD ■^-i CD ^— 1 CD in |\- CM L>J TO >Ti VJ r— • CO ! I ! -*—' cn CD u_> kO i\- Tn- 1 ."» i "•— m t-H m en CM IS in CD C\] LD cn TT 03 ro CTi \T CD IV *—* t-*- t" : ~! -r-H *r!- CCi ^-h m CO C\! i. f: U I IM |\_ ir. in Ln \f ro IT) CM *-> T _. ijj t. "1 IT"! UJ U3 'Xt 'Xi 1X1 l._T| kD Uj 'X U? '- in U." CD S S3 CD CD r- ID r- m r- ro ro CD 'S s m d o j ro r< in in 133 CO — i •u U- i~ 1 s=) □ U- en o Q UJ '■■' z> '-■ UJ ►— * x Ok h- h- 01 !jJ LU >— • >_ 01 Ltl : h-O >— LL •cn :z co co o > dJ r^ -Q CO 3' G3 i3 '3 iS iS iS 3i IS © iS iS '31' i3 iS '3Zi IS iSD iS 3 CD 3 i3 3' y- O UJ Li J IjJ UJ IjJ UJ LU UJ LU UI LU UJ UJ t—i ro 3 •r-l IV .— i in 01 LTj ro •r-! CO JV) ro _J ■r-l ro ro l"T'l .—i in 3 in IV •^r ro -J ro _J ro cn CO cn •<* '3 en en T— 1 LT» 3' 'J3 ro r-L. ro in rv CD IV ',D in C-i l?l ":j" CTi y IV ^r 13 i.£i CM iD "tf 3 'ati OJ rv CO Op LO LP in ■^ "3 ro P0 ro CM CM -,—i ■r-H IS O iv |V rv IV iv rv IV r-- r^_ IV IV IV IV CM CTi Ul iTl Ui CT| iXl cn Ui CTl i.Ti 0"i cn iT'i CM '3 3 i3 'S 3 3 i3 3 3 iS iS •S IS 43 IS 3 3 IS 'S 3 3' iS 'S iS ^r" 5 UJ LU UJ UJ UJ LU UJ UJ UJ UJ LU UJ UJ t_ # 3 iS 3 3 '3 3 CJ 13 3' '5 iS 3 IS _j IS i3 3 13 3' i3 IS 13 3' 3 iS IS 3 _i 3 13 iS iS i3 3 3 3 3' iS IS 13 3 is 13 3 3 iS 3 3i *3 IS 13 3i i3 IS UJ 7~ IS (3 i3 iS 3' i3 iS 13 3 3 'S 3 iS r-j 3' i3 3 i3 'Si iS ■ 3 13 3 3 iS iS 3i *? IS 13 iS 13 '3 '3 13 ijj '3 3 IjJ '3 iS to "-" *-" "-* ,-w •—i .—. T-* "- 1 *-■ ■""i - -1 '*-' . - r ~ ! UJ _j 3 13 i" 1 '3 '3 3 i3 '3 3 'S iS 3 3 LL -' UJ — " __l l_j UJ L . _, ^_, •r-3 .^_, ,-i _, ^, ^, ,-! ._, , rf ft 3 13' iS 3 'S 3 3 IS 3" 'S 'S iS iS UJ 2- LU LU UJ LU UJ UJ UJ LU UJ UJ UJ UJ IjJ |T"I 31 3 1 3* 3 i3 3 3 iS 3 iS '3 3 3 iS 3 3 y~ _J 3 13 13 3 3 3 13 iS •3 !S 3i 3 1T1 QC _J G) (3 i3 '3 iS >S '3 iS 3 iS IS iS IS iS IS l3 3 <3 i3 IS IS S 3 3 3 iS 51 © ■3 iS iS i3 iS 3 ■3 3 i'J IS iS iS 3 '3 i3 i3 3 3 3 3' 3 iS 3 iS 3 ■^-4 ~ "-* "-• ■"- 1 ■— ' "- , •"• ■—' ^ '-' ' ' *-■ *-' i3Zi 3 3 CD 3 iS E 3 iS iS S3 01 o h- u co in ro ■s IV LTJ CM '■D IV CM 'JJ CO (V iS 10 IX) 3 Ll. o oj r-- in 5- rvj *-H cn 05 [V LT.i \r ro © iJi iTi iT. >ri ui cn CO iX' iS CO co 03 -J .r-H 3 'S 3 3 3 3 iS 3 3 © iS 3 ■a: u UJ Qi LU i— cc ^ J5 }£ V* *« •K x«> N *»ta ^■? X *e K 01 IS' IS IS CD 3 3 3 'S 'S 3 iS iS i^i IS in 13 in iS in iS m iS in 3 in 3 a IS 3 rH -^ CM CM ro ro tr ^r in in UJ en L£ UJ 134 tSl a _i LU • — i U_ X ,_^ h- >- w LU 1 ' X IX 1— H Qi LU L±J ■— ' 2_ Cfc. x - iTi OJ tn p_ CQ LD L^.' CM G3 i. :) cn iTi LD ITi CM CD •■i CD CJ in 03 o CM ^ LD CO cn CD r- ro cn 'X' C\! CO in »-* rv ro iT: LI CM — « *— < i v - CM LO cn iM M r-- cn E 1 in ro *-H CD CO u? i.M 03 *tf — i IV ro CM ■—* cn rn r-- in M* Ui X; 03 03 03 01 03 ; J3 I E I E i L'~ i OJ ! ro : r ~ i 03 I cn I i i iS iTi CD E r t E CD E' E |T| E CD E E Ei E Ei Ei E E Ei E E' E 1 Ei E III LU LU UJ LU LU LU LU LU UJ UJ UJ !.U 03 CM LO T r-- CJ IV *— i E CJ i.J3 03 i.£i iJl in ro \| r^ ro E E' *— ■ ro '43 C3 LD T p.. iji LT.i CD in tTi CM 'X in LP !J"' ro in OJ UJ in CJ 03 ^r ^H K. ro 0"i u: ^-i p.i OJ *-1 ■r-1 ■«— i E E E Ul cn 03 C"! 03 m 03 03 03 CO CCI 03 03 r-- r-- rv rv ^- iTi lT| cn U. cn cn cn Xi X\ Xi cn cn ■0"'i Xl Xl Xl iTi iji iTi ijt cn X'' o-i cn Xi cn in i J? >! j-; }•* v-: >* }-C x<- J'; 1 E E E E E E' E; E UJ i E in E in E in 'i.i U .' E 1 135 i.n « _] LLl t—t it- er .^ \, •\_ cc _J Ci y a Qi a -- LL a i=) Cfl j — i — i .J / 3 ■LL CD uJ r Ul Ul Ct £ t- LL LU Ul •— . s~ ii_ ill ■! z: u. LL. Ql c l_l _' to T" LLl Ci X 1— ,Ct — ' U- •a: 2: L0 tr O •! » — i U Ul LX D. 4 u r^ CD r— -Q rC ro LU _! r_i UJ UJ .-r 21" li 33 CD 'SI Ei E' Ei Ei CSD E' Ei CSD CO Ei Ei 'A.' CI Ei E' E' EJ E' E' Ei CD ffj Ei Ei Ei 2: i__i UJ UJ LU LU LU LU LU LU UJ UJ LLl UJ UJ i — i *—i CD IV TT 'X ro CD 1T1 1X1 T Ul CO O- _i *— 1 *T iv CM CM .— • Ei •vT in IV ro OJ .j ro Tt Ei CM cn OJ f— < 'Xi QD r- T$ CO E) 1 — ! CM CD LO OJ a^ CD IV UD 'Xi r- 1X1 ■^-1 in LTJ xr -3- 1 ro ro ro ro ro ro ro t)" \t CM CTl LD ro CSD |V ^j* ~^ CD U"l CM cn 'Xi CD CD IV IV r- fsl IX '■£> '■X LO LO in xr '■\r CM CM i?J CM OJ OJ CM OJ CM CM OJ OJ OJ OJ '- ' J Gj ,_ ^_ ^ Ei ,_ ,_ _. _ ._ ^ Gj T .. IS CQ Ei Ei Ei CO Ei Ei Ei Ei Ei CO El Ei Ei Ej CD CD E) CO Ei ro ro Ei o LU LU LU LU UJ LU LU LU LU LU LU UJ UJ 1— < m i—i Ei to »-) IXi '£1 CVJ CM CM -■— < ro CVJ _J cm Oj ■^r r_ in *X CM 'X CM IV on ■>-H ro _J Ln LO xr CD "X m ro IV UJ xr 'Xi X r ■D 1 — 1 ^r *-H h- OJ in IV cn IV Lfi CM IV t"t ro UJ T* ui ro LD CSD f 'XI en OJ LT ro co ro LO N 'X.i 'XI LT' LTJ ^ ro OJ CM ^h CO ro ■ 1 1 CO "— * "^i" LH lh IX LO LO LlO IXi LP LO LO 10 M- 'sT OD CO 'Xi CTl 1X1 cn 1 cn 1X1 •Xi 1T1 G'l 1X1 ro LU CO CO E' E' CSD CD Ei 'D ro EJ ro Ej i — i LL. El El CS) CS CSD E ro 13 Ei ro m ro Ei Ei E' CSD CD CD E Ei .ro CD ro ro 2; UJ LU U LU LU LU UJ UJ UJ LU Ixl LU UJ o Ul CD CM ro ro ro LO on *^r T— < OJ LO ro t— 1 CSD LO CD (V OJ OJ 'X' ro ro ttI- LTi LO "Or _J El CD IV ■^H CV! Ul OJ OJ OJ co CD OJ OJ _J Ifl Ei ro fs- CM in iv od Ei Ei «-H «— 1 t— i r-- !•£' 5T CV1 — -i Ul r- in ro Ovj 'Ei ro ro .V ui CTl 1X1 1T1 en CO OD ro OD ro ro N 0- CD CD OD CO OD ro CD ro OJ CD CD on 0~i T ^ Ul CTl CTl U'i ro Ul cn cn CTi ro 1X1 Ul ro CD El Ej lit' ID " CO E O Ei ro ro E' ill -■ -- — — " a l~ CJ 13 ,-, ro rv cvj Ul IV <•£■ IV tT or. ro : &: 5.* ?^ >? J? X 5£ J,* y 5^ •>c K ro CSD CSD E> CS CO CD iD CD E ro ID ro cc IXi E> in CD in E in •D LO' ro in E» E -^ -H CM CM ro ro "sT xr U~i in ro CK CH LU 136 CO LU o LL I- CD UJ A- LL 1- o o LLi '-'.- -._," Lu ' — 1 X Ct: h- h- G£ LU LU •— ' j_ Ci- LL. Ct T. LL U, Qi i_J CD i 0) >' LU ox '— LU — i »-l ■*-! ■—* •-> T-" — ' »-• *-■ ^r o h- CJ •I u_ o UJ •I 0£ XX. O G£ IX LU CD CD CD UJ LU LU LU Ii 1 m in r- CM CTi ro IV; ^r 1 1 CM CD cn ro en LD ro i "t t 1 - ID *-l m CD M uj CM 1.1" m \] CM CD Ul rv GO CD i-Li CD r- rv IX: -i -T ^r 'Xi CM in ^T r-_ i\- CD CD 03 CD CD CD CD CD CD CD CiJ CD CD CD CD CD Hi LU UJ UJ LU TO CD CO - - LL ■cn ^-1 LO 'j ■ t- ft i 7^ LU 3 d ! to 1 ! LU X f_J Q h LU Jt_ t-H Q Cj LU .£_ LU h- - in LU i—i a_ 03 § U. U_ i.i. □ o en z LU a X i- ft i — ' 111 •:j- If) cn !— \ •I Ul CD Oj LU LLi a. O ii: il- UJ CO & III Hi rn rv LTi n,i iv iT'i CD -vT PO Ui 'Xi CO 'Si' CD CD E i— i CD co Ei CD LU UJ UJ LLI UJ UJ LU UJ LU LU '-ill ID CVJ CM in E (3 OJ 'X' ro ro 'D 'X -r--l CM Ei CD 0"i cn r- i o CTi L-l IJ"' r-w a. E ■r--1 f'.J ro — . ■^r in 11 CM r\- i—i CM T~- ITi \t Lfi iv CO "X 00 V E' IV "X en i 0~'i 'ZI' E. CO ■— i CM OJ 10 cn CD r-- !■■_ 'Xi m T [<-, CM *— ' id <■£' IX '■J_i >.n ID 'X' -" *- 1 IT: 'Xi 'N tn f^- r- 'X' IO OJ tr rv CTi — ro t!" U"i *■- 03 ijn tTl ■»-< »-i cn LTi *$ \j ro r*j «— < iri iTi 'Ti 03 IV ID ■-r y Xi" j XT xr ^r ij ro K) ro ro ro CTi Xi |J i cn cn ij'i ui CTi IJ ■' cn ITS Ul ui 'll' CD E CD E Ei Ei E is: Ei E E E Lt — l H LS CD i--i I'' l\- OJ cn r^ ^X : IV E "\T cn CD ■X E L""' lT" r LO ^-H 'Xi C'J 133 ■^" --* h- ro |T"| LL Ei CO r-'- Ul tt 0-J -^^ iji CO r>- LH 'NJ ro E ■ r ': iTt Ui cn O"'! 0". > 1 1 1 1 1 m 133 CD n3 _J ^ E CO c- E E i^j E E E E CD E 3** 51* X >? a E E 33 E E E '.! -i E in E E Ei — -i CVJ •>; X J5 v:- E CO E CD E in E LO ro ro tf E E in e 138 t/J UJ CE o LL W to LU CK LU > LU LU ■X 01 LU u. a: O D co LU 1 UJ D X cn cn ^ a: o z: en LU D_ CL .a i — i 1 — 1 i33 G3 id CD CD CD CD GD •Xi CD CD CD >— J LU LU UJ LU UJ LU UJ LU UJ LU UJ LU LU t— : -r-l IN- ro CD CM CM CD CD 1— 1 ro cn CD CM ' © KS tt- IDO LO Ln TJ- ■\T CM CD m 1 1 i IV _i iXl ^— ! f\_ iv ( J T— , "\T ■T-H J IV LT: IV ro 1 1 :j"l I 'J i 0: \T cn ! r ~' IN- ■*—. rrH f'~ ££i o • U"3 OJ 'XI IT} m CD N- TO -r— i cn r- IV 10 IV •^-; ro ID CD ■•— * ro 10 CTl i — i "Vt- IV ID ro l" r ~l ^ ro OJ ■*--. f—t i3 cn GD CD r-- I.J- l,j- in CM LTJ in in m U0 Ln ■"* <3- kt t \T TTt- ■\t G3 CD CD CD D CD (S3 CE3 o CD CD g? CD ll. a h- u -J a u LU ■T IX D£ O OL D£ uJ CD CD © is a IS CD LU UJ UJ LU IU LLl UJ LU LU UJ LU UJ HI in Li .' C\! gri CM l~'J LO ro CD ro \r" CD ro ■■—i !V cn nj OJ ID IV ij-i •.— • ■*-h T~ ! LP cn Gj cm — * ro CD 10 IV CD' m r— i UJ in CD CM ro CD ro CM 133 cd 'i ro T CO ro i.n or, IT* CM ro ^r ■\T in ^r ■\t r" CM '2Z' cn ID *^ -.-i CD LTi CM ui 10 ro i ■ r-- ■^j" LO LT LO in ..._ Tf ■ ' ro ro ro ro CM ■?0 IV i%_ v r>- IV IV r,_ |\ jv r-- IV f-_ r-_ i3D CD © o ID CO CSJ LTJ O (S |Tj IS? Ei ID >D © '-^ -*■ ^ *~ -■- -1 " "*"' "*" - u -^ IU IU IU IU IU LU LL' LU LU IU LU LU III in IJj CM LO LO "T r«- *-i OJ ro UJ CD in CD LO 'JJ |V W- i.n m CTi CM CO 03 CM IV CD cn \t ro i < J ro ro CM ^--T ID OJ in •Tt- IV "* t^L; ^ 'D CO v 'D ^r in UI OJ r<- ■\t ■xt ■^J" ro OJ *"< i3 IT, IV u:i ro -^H cn IV '•D in T 10 OJ *-< cn m IV ijD in 10 r\l CM CM CM CM CM CM *— i ^H r— •- r-* i~i ■^-i cn cn Cl 'T'i cn cn CTi cn x ;/; ;.;• ii K n 3*4 *i CD iD iD ID iD (D iD ID iD in iD in iD in CD in ID ID LO IS -h ,-. ro ro is ID in iD iD in is m 139 LU ll I- CX O 03 a> X3 1 i 1 ! ! I - ! i ixi UJ Ijj 1 -- 1 ro '<$ 0*J I _J i iv -r— * -r-i i i 1 M m 'sf I...U Tj- '.£ .' z_ i 13 03 1X5 i i 'X' i ! -l 03 rri 03 ! CM 1 CM CM CM 3 IS 3 s El CD 1 1 m IJi IV *-: l"0 IV CJl !.l 1 ""J- in ro ro LP 00 en UJ ■^r CO CM -,-h r-. •! 1 f'O l'"0 i — i IV * ! C'J -> — ! p •: Ln 0-J ro m i. n IT) m IX m ■st ro ro ro ro <3 IX! IV U ': iTi 3 'X CM CG ■*f IJJ Ljj l. '--J js r-- j\_ CD CD IX! IX' IX ^ 'vj CM CM C'J CSi CM CM CM CM C'J CM 1 1 CD 13 IJJ i3j CD 3 UJ CD is 3 CD £J 13 1 I ! 13 )3 (S3 CD '3D 33 iU CD 133 <3 'X' CD <3 i 1 O I UJ LU UJ UJ UJ IX UJ LU UJ UJ UJ LU IX 1 Li'l in ro CM CM '—I CM CM 1 ' '1 'X U'l ^T i i -J i 1 ! 1 03 r... 03 CO Ln 03 o-j iv iji CTi Ln ro 13 CD <3 ro en Ln ij '- Ln i LlJ 1 \\ i i 11 l 1 1 IX) CF'i iji 03 U3 CM 'X iji Ln 0-1 CM |V IV IV 0J CD X ro CM I-'; CO 1 '— ' ■>J ^— ; fTi IS iTl Ui iXi IjJ 03 ijj r.. 1 03 CD CO 'X CO 'X 'X i.n L r j IX I.O IX IX Ll) CD 13 i3 CO I i—i ! I.X X i U 1 0c 1— ! XI 1 rzz i £3 I o <— i 1 111 j UJ 1 UJ i 'X 1 1 ! "V i ~ I a: h- 1 Od 1 h fX LU 1 3 1 o. i IJJ 1 Qi 03 'X 1 21 1 i IX |J_ CL' ! i o i_i _i 1 03 1 i i T UJ 1 i O _X. 1" — i— "i 1 ! »— » X ■x :z: i | to x « 'i i 1 i— i i i CXI i i LU i i in UJ LU UJ IX UJ IX IX UJ IX IX IX IX CD IJ r IM T i Pl iT: *— < ■>J ■>-H iXj 03 r.i CD *tf \i IX I""' J iTl t r. Ln IJ! IV 03 IS C'J CM i3 'X IJI U'l 03 CD ro CM 03 CM CM >Xi ro m 13 01 LTi Q3 ijj-i CM ro Ln I.I 1 •XI r*- r-- IV ''JJ iTl r- Ln T i"M 13 IJ 1 l£i 0J ijii_i 1 ) 1 s.jj ijj in j.j j.<- ><• K JS K X X> i3 i3 i3 3 iS '3 Cii 'X' UJ iS 3 '3 in GO LO 13 LH 3 \i: 'S !'i IS IX ii.i -. OJ m (0 140 o 'V LU Or h- Di LU UJ ' ! > u: 03 -I z L^ u_ o: VJ < ' — • en s LU o X &— O •— ijj cs cs 133 in E< in *-i IS cn iv in ' T 3" CS 1T1 U- Ui Ui Ui iTl \T — i IV f .■: U'I iTl iv LD ■Sj iX- C l!' | ; LU LU LU LU 1 I ! LU lii LU LU L'J L ro Lf'J CM "3" V LO r rv £~i LO iTl ^ 01 LO ■■— i QO LO ro H-J t- ^-. CS iVt '.£ in ro CM ■^ !T. | r>_ •■J3 L f\| en in CM Ui CO ro |T*I CU ro iXi rv. ^ iVj iTi CTi CO CD 03 03 03 rv r^- T- ■J3 L h'j i"..j Cm CM rvj CM CM iM CM 0-: CM Cm r in l^i CS CS i3 '3D IS li. 1 ITl Ei CS Ei Ijj CS IS '—' - v ~ - 1 - -^ i- — — -^ -^ -"- -'- III LU IU UJ UJ LU LU LU LU UJ L^J UJ IU IV LD IV LU 03 Ui ro i-< ro LD C"'i U'i ro 0,1 ■£. -T IV ro CM L0 i^i r-- CD 03 CM cs ^-i ro '•J ro T-t r- •»— i \r \T IO> Ei LD Ei *-H — i CD CM ro IT". l£i '.£1 TTf U) ■^H Ui IJ3 ■^r • T 1 ! ro 03 iM 'J3 Ui CM in IV cs ■•— i ro 'M ij3 IV 10 i3Z' i n l?-l Ui LO 1-H 03 ■^r ii-i m ro CM CM OJ t— i -^-1 CD E' Ei IT'I Ui 1 1 1 '■J3 IJD '•j_i LD LU '.£i i£i C£i LD LD LO LO LO Ei E E LO E E IS LO E E' |V| i£' Ei LTi E' E. Ui cs E in Ei El Ei -. — CM CM ro ro ^ ■^r in in LD 141 en ■T •-, h- > ■T . i O z D tv o > Ll •I CD Cj U.' ill X 0£ h- J- Cj o UJ i_ >■ ••v UJ 1 — ' X a; 1- h- Qi LU UJ •— ■ H cn. cn a: T. LL. U_ Qi CO 2: uj axi-o co en q -x i — i tj UJ a. oj ll. s CD Ll o _1 UJ - CM CTi iT": in •S3 in ,— * ID CD iX IV in ■tf CM CD en 01 Cl CTl CTi — co CD CD © iS iS CD Jl? >i x *£ \* >« j* J4 IS) IS CD CD CD CD IS CO CO in CD in CD in co in IS ro m 142 '.D I LU \r cx LO W> b") iS E CD Ei 'S E< E' I 1 ." 1 ! 'S IS CD E iS CO CD IS O CD \\j CD E' iS CI' 'S CD CD III LU LU Lu LU LU IX LU LU IX IX IU UJ FV CD IJ'i 01 iS »-) cn \r 01 \r r-_ *— 1 !\- r.i \~n |V CM 'X Ui ro CM IX 0~i CTl IAJ l\* fM Vr IX CD cn E' CM ■tf 'X lj"l IX CTi LO IV ITi iS ro IV ro iT' 'X xr ro *\T in CO ro CD \f cn ■S} IS in l-H r-- ff: Ui in *~i iS r^ in CM iS IX' m IX 'xi IJ.i in IX *— < *-4 e CD s CD CTi cn CTi cn CD CD CD iX CM CM CM CM cm *-< *— *-< *-< '-I *-> — < *-< IS CD iS s Ei Ei iS Ei iTl E E' CO E CD Ei iS (S3 IS CO CO IS s CO CO CO CO III LU UJ UJ UJ UJ UJ LU UJ IU UJ LU UJ cm ~- « m CTi CD CD in CM r- T~ o ct □ _>- Li- <£ o en f- a _j _j ■•-^ LU •— • X Lil 1 — H Ci LU LJ -— i T QC 03 — jj 3 CD G3 S S3 G3 ; ? G< © UJ LU UJ UJ LU LU UJ LU LU U! U.I Lu UJ CTi LD T iTi Ui IT, UJ ■»— i T~^ -r~, T--T .—i IV |V cn rv iXi CD IS LO ro LD CTi CTi 1 1 1 r^- sv CD LD -•— < ■«-* Ui CM CM *— ( C£j CTi |\- i<) TO '.jj ^— < CO i. i ! Li > [V CO CD G? iv ■r--! ■•— * cn CM LD 01 ro r>- t-H LO -j- T ~! co Ou u i— i S: on LO LD l"0 CM IS 'S LP 'J: ~r Ui «-» 1 |V} S3 G3 Ei G3 G3 0"'! fv LO xr CM * ' "= ' " *~* 1-1 ^~* *"-' * —t CTi Ul G' iS iS G2 G? Gi G) IS IS 'S 'S 'S LU LU LU LU LU 1 1 1 Ul LU UJ LU L_J UJ LLi S3 cn •^ '■D CM 'S ul cn ro 'S r-H ro ^r LO r- rv T iS CTi \T LD M" LO "-- ! Cu 'S S3 cr. i Q ■^H ro ro *-H LO cn Ui IV .,--* ■tf ID m o~j LO CD ITi LO CD Qj LO cn IS fv LP ^r ro ro ro ^r LO CD --^ LTi Ul LO G' fTt CM ■^j - ClO UJ iS CM ^r IV cn *— 1 '"l i^- CD IV LD LO ^r M" PO CM -r-H 'S S3 i 1 ! 1 LTi IJ-| LO LTi LO LI "J in uo LO IS LO TJ 'Tt S3 S3 in S iS Gi iS iS LU LU LU LU . w UJ LU LU LU LU LLI •±i LU CM UJ 00 in 'S f\- ^^ cn IV CJ UO r--- CTi LO CD i\t ro 1 1 i L0 ■^-H C-J ro LO iTi r^- *—i *-« ^ LU ■•— * Tf CD ro ro v* T— 1 cn Ui *— i cn *-i Xj" cn ^r S UJ •jo LD fv. 1 1 1 '-i Li CO in *-\ r- t t— r- M- '—I 03 uo ro liJ f>- LTi ro iS iTi LO ro •— < CTi 'jj •vT CM :X"i CTl cn U'l a. CD; CD CD 03 IV !'•- r-- rv rv T "~' *" * V— i 1 ' ^^ ^ ' T ~ l ■" ' *~ H '~" < ' ' iS IS iS iS IS S3 'S iS 10 Gi U0 -■ c-J 143 l_j K±.i &> — ro IV CM u, IV LD i - 1^.1 Tl- cn UO s UO re? in ^—1 LD CM 00 ^r ^—1 IV ro Si "S CD IV uo V CM rr~ 1 cn CD r- in t ro iS Ul cn Ul J' cn Ui 00 00 CD LU UJ LU — IS >3 'S IS - > ■w' UJ i — i I Di h- h- Lt LU LU i— . * 3i CO. 'I \i- LL |J_ IV □ O ' CD ~t_ LU O X h- O »-" Li. ■I Z CO 0" O •! 1 — I u LU DC Cl n3 co <3J JQ n3 U LU UJ ? ,-■ Si ?£ J£ ?■: $£ Q£ E Ei E CD Ei E E E S3 E in 'E' in S3 in S3 in E E — -h ■tf Ei E E E Ul Ei LO Ul Ul E LU E E E E E S3 E E V iTi U3 —i IV 1-1 E iv ui rr pi CD CD OJ CD E CD 144 LO LU > - _> ■■— ' LU •— i X a: h- 1— Q£ UJ LU 1 — i T LY m a z: Ll_ Li Qi Cj i_) _J rz. ^P — Lu ■a: z CO 01 Q ■:! »— t LJ LU Ct cl -Q 00 O) -Q fO 1 i CO CT. CD o CO CO 'D iS iS IS iS 1 1 CO CD S E CO CO CO CO CVl co iS Si iS 1 1 o i Lu LU LU UJ UJ LU LU LU LU LU LU LU UJ i 1 ' — ' ! P. 1 LTi CD LO ro 03 CTi LD T ■»-H in — H IT 1 i _J ! V IV r- •^-i 'D ■^-t "\T CD *— i l?sl Ui CM 03 1 i _J i ro CVJ iv CD \t IT, •— i CM cn Gl LT, KQ CD ! • •— ' i cM IS IV r+ ^-) r-- ^ 01' |M rv CD ro U3 1 1 ±z ! rvj CD [\_ lo rn CD •DO m ro CO CO m CM 1 i LH LO ■^j" TT ■si ■^r ro ro ro ro CM CM CM 1 1 GJ i CD CD 03 CD CD 00 CD CD CD CD 03 03 03 1 i rsi i ui •Ti IT iT iT cn on CTl iT cn Ui Ui cn i i : . \i |Vi CD i3Z- CO is CD iS Gj IS CD CD 1 i co CD CO CO CO CO iS iS CD 'S CO (Y"l 110 ! X 1 i i ! LU LU LU UJ UJ UJ IU LU LU LU LU UJ LU 1 1 ' — ' 1 CM fV LD ^r LT3 IV ~-* LO IS 1. o CM 03 ro i 1 ! 1 !"•- ■lj • cn CD l£i ^r iS 'JJ >_j -Ti- 03 1 I _J i LO T-1 IV ro IJ! Ul ■r-< IV CM CD ro CO ro i ! i — ' i in LO *tf f ro r-:i n CM CM T— ■ -^ CD S 1 1 LU T" i IV |V rv r-- IV r^- IV r-- |V rv. IV IV r- i i l\i 1 Cl U1 Ui cn cn iT iT iT en cn cn 'J i Ui 1 ! ' — ! ■tf i Ul iT IT iT cn on 'T iT iT cn iji U: •Ti 1 i cn i IT. iT cn ij'i iT cn iT. •J'l iT Ui iT'i Ui on i 1 LJ 1 GZi CD CO CO i.D iD iXi iS iS iS CO CD iS i f •— 1 U " CM ._. UJ _l 1-1 LU CO iE' CO CO CO CO CO IS CD tS IS IS iS 1 O S3 CD CO iS 'AJ CO CO CO iS CD 'S CD iS 1 LU -r LU LU LU UJ LU LU UJ UJ LU UJ UJ LU IJJ I JL_ o CD kO CO CM i?-i --— " IV CM 1X3 UJ IT OJ '-JJ ! -' X ~ii X >•? S4 X a 1'. J.< a J;> X I QC CD CD Gj CO 'D CO iD IS ID lT"i IS 1 T " si i CD LO CO tjT! iD ID iS u~> CO if/ Si l?J CD 1 0_ a a: CD iD — ■ "- 1 i"i CM f-J ro ■^r ^ IJ-i ITi |j3 : LU 145 UJ Li. *. CO 1- ! ■y~ 1 r ■ _J CO Lu * Ql Q-: G LU L^ 1 J~ UJ •— ■ X 1 1 1- i ck id Xj i — ■ j^_ CK ixi " - UJ _j i CO (S CD CO G3 CO CD iS CO CD G i CO co IS ■ 4 " -*- "- 1 - w_ -1 " — "*■ -*-' ~- -'-•.' 'x' UJ l_Ll UJ UJ UJ UJ UJ UJ UJ UJ LiJ UJ UJ 1 03 "\T Xf XT CM *— < UJ •ST CM XT CM 1, ! KD IT. XT 03 CO i3ZIi fTl ijD ro r-- U"i CM Ni" i >. -j IV. CD ro ro CD 1. l"i r-- CM n\ iTi .,_- X 1 I'M f\: .._., iTy in CTl CM ro ro ix U3 ,— i f": ! XT MS o:i U'l ■^i CM "■.! LD Uj rv. i"-- CD IX' 1 "xT po C\ ! ■r— 1 *-H CO CTl n~i r-- U3 lo Xt ro i Xf" XT xr XT xr ■-f rO ro ro ro ro ro ro ; 01 CT) i CTi Q'; CO CD IT, U'l iXi U"l en IT', , Si CD © CO iVi iS IS CO CO CD CD CO CD i 'ID fUZ' 'Sj UJ UJ UJ UJ LU UJ CTi ro Uj vr CO CD CTl t— i - i ^T LP pn CM r- Cxi SJ [■- IS l.£; xr ro -^H a- on IJ'I 01 iT'j iTi CG 03 iTi UJ ITi 03 od 03 iTi CXI IX. cn CO CTl CTi J', U l IX: ' IJ 1 iTi iXj CD CO OJ 1 ' ' iX 0~i IJ'. IX' CO IS CD; CD ixi i C J I'D CD CD ! UJ UJ UJ UJ i CO *— < CM iXi 1 U': CO 1 ■ " 'i XT 1 r- 03 1X1 IX. 1 1 . , CO U3 XT 1 03 r- [\_ IV i ij3 oj 0~ CTi i »Xi iXi CTi cn i •s iS CD iS CO CD 1X1 Cxi CO IX CD IXi CO ! UJ UJ UJ UJ UJ LU LD LU IX' u. 1 1 Ixl uj i t— : 03 CO IXI ro 03 LTJ r. 1 *— 1 T-H ■^H LP r- iXi •^-i ^^ ixi ij'i i-- x* iXi Li" iX'i CM 1 XT iS 1 D r-i . -j UO IX' i n f'J 03 ro iVi U3 i.n LD uo "^i" XT "X ro ro r-j OJ •—f *~ : i 1'-- IV. IV. i- r^- IV fv r- r-- IV rv 1X1 'Xi Ul iXi X'l :Jl U'l iX'i iXi _ in CO Lf) 0~i r- LTi ui U l >Xt iS CO a TT '■D iv i'X 1 xr IXi 1.0 i Hi tr «— i N ro CO 1 cn IJLI r-- in ^r ro i CO CO 03 1 : . 03 03 1 CSJ CO CD CD CD I CO iS U3 CD CD iS CO L0 CO Ln CO IXI l-J — CM cm ro ro -tf 146 CO UJ LU i — ■ X CK H 1- IX LU LU •— • SI £K CD 1 Z." LL. ll ex C"i O 3 CD ZL LU O J •. h- d ~ LU -I — co cn a - LO ro *-H IT) IV uo ro in LP in Ln T *tf s iS iS CD IS CO iS CD LU UJ UJ UJ IJJ LU UJ UJ UJ IJJ !.!J UJ I.H IT, 00 m •S3 r>- IV LO IjJ CD r- <3- CO 'S -tf ro ro ^r •x IX! CD IV IX LO IV) en T V r : CO en •- * IM CM t— ( en i. .-> CM ■X CD CM CO ro oo xr iTi -Q CD i\ ' r .r •.~! fv_ 0"i IV CO LO ^f OJ --— 1 'S cn IV 'X in ro l"C| CM CM CM UJ CM C'-J CM *-H -r— ^--i ■r— : ■•—* en cn Ui ui cn cn iT' en cn cn Ui CTi U: iTi 01 CTi Ui en iji O'i U) Ui Ui Ui Ui Ui CT, ITi Ui iJi Ui cn OO cn Ui U", iji Li Ui iS iS CD CD CD — iT' m IS 1.0 ^ I.O 0.) GO g- —1 jV. 10 iv iS CD r- LO 't UJ *-< Ui CD r-- LO "t ro iS cn Ui Ui cn Ui Oi CO iS 'X Oj CO 00 s if) cr O! ! - i^ a a i— i jt- u •I o u_ n r. 1T1 h— |— ■y: 1 n 3 Hi LU V Q£ Lti — ' LU i. 7" co en i~A .3; ►—i u Li! D; DL ai 00 a> CD iS a iS 'S IS cc 1;- iS 'S 13 IS s '3 CD iS IS iS iS 'S GJ Xl Ul iT'i cn CTi Ul Ul Ul lTi Ui Ui cn Ul CTi 'Ti cn cr. CTi Ul Ul Ul Ji Ul Ul ijl IJi ■ LU c"T . CO CO fij £■ ro CD ■^r CM CM Q"i CO Csi ^— i 1*0 U1 cn CM CO '.n IV ijj CO LO cn ' LO |S_ <£i CM ro T-") LO ' .U VD Lpi po ~~: cn ! £i ro CD LD CM [V CM CO LO CM cn m C\i cn '£■ CM CT] Ll ■ CM l\*i C\3 CD CH cn cn :" CO CO IV. r\- |\- rv r- V- 'X' LC '£ LC '£• '■- UD '.j_i '.J-i CD CD CD iS TS LU LU LU 1 1 1 LU UJ CS" [V 1*0 LO LO LU \i CD U.i LO 1 — > ; . \ T— ) Ui Li j i j-j T— i LD "* CM j rr< ro c<\ !ji n CO U"i CM Qj m CM IJ1 © fT' cn i i !J| CO CO CO r-- r\_ IV p. 0"1 CTi iji a % i U'l 01 cn U: ij-i IT] cn IJ 1 'S iS - CM UI IV i.JJ fv IS v" cr-i s LTi 'S LO »"H Ifl OJ CT '^" ■*■■* r- ro cs Oj fv if"' T iM r~< iTi iS r-._ w> "\T Ei UI iTi cn Iji cn iTi Oj Oj i£' CO CO s CD iS CS iS in CS 'S LTi iS iS LO iS iS ID 'S iS LP CO IS ~H TH OJ OJ ro ro X!" -Tt in LT; iS i.L! IV. ! CM ! iS i CD Ui 149 LO o _J UJ u. .•~. 0- a ■X h- i-j CL LU co IS iS 'S CO iS 6" iS Ei iS iS iS iS OB iS IS CO iSi 'S X: IS s iS iS 'S iS LU LU LU LU LU LU LU LU UJ UJ LU LU UJ 'Xi LO LO ro :-- ro IV IV ro ro -rj- CO cn CM 00 CD cn LO ro <— ' r\i CO CO rl CM i.£i [V LO cn LO cn IV 'S CD *—» *— , LQ IV rr OJ CM cn ■■■■f LO 'Xi Tf O'l *—H i — Cl ro {£} i" i r^ m 03 'S I'M T^- LO IV 0"' CD cr. XT\ 'J 'X' UD CM cn m ■^1 rv ro cn UO i—i |V Uj LO UO LO -cr \t \r ro ro CM CM CM .— ■ UO LO LO LO L"l LO LO uo X' uo LO ID in o Ei G: C3 S3 I*!' CO CD CD iS CO Ei C2 CD iS Q CD CO CO €0 CO CO CO IS CO CO CD co CSD CD co G3 CD CO CO (S3 CD CO CD UJ LU LU LU LU LLl LLl UJ LU UJ yj LU LL! CD U, 0j on L0 -cr CO CO CM en t F- ro 'Xi "X *— i L0 CD LO CO -i ro CM CM co ' D n ^— t »-i io r\- ■tf LO ■"t n. UO CD LO OJ cm cxj *-i 'J: IV Ln CM cr uo 1—1 UD 1--, Xi CM co CO Ln ro --1 cn 'X -yj* CM cn r - "J LD x Ln Xi uo in T *-]■ id- --r ro ro ro CO CO on CD CO OJ CO CD CD CO CO CD CD cn IT, Xi cn u, cn IT, iTi a, cn cn LT'; cn H'| (S iS CO CO CD i % , (S3 IJ s CO iS CO X Ei IS iS CO s CD CO CD CO CO CO Xi iS Ei LU LLl LU LU LU IJJ LU LU UJ UJ LU UJ UI OJ LD ■T ro .' CD r- LO CM CM ro 0-J "tf ro ro ro CM Xi I.X i—i <-r m ro Oj CO CO IX LO ro Cn " IV CO i—i —< C'. rv Tt" on CD ro 00 CM fv. — 1 LO IS •nT - LO M- CM Xi cn cn cn I" J Xi X -tf •^H rv ro s iTi rri r>- in '.' f. CO CO CD CD Xi u ■ CO CO CO CO >- X' ?4 7i j.; JC 3»* %£ K J; J iS CO CD CO iS CO iS CO Ei CO LO E 1 UO iS UO E- UO 'S X iS Xi — — 150 CO a _j LU » — ; ll ..-^ > X X: iX' CO CD IX iX iX ro ro *— 1 CD ro IV lT, iTi TJ IJD CT'l ■i— * "Vf i.n in X »-< ro '\1 in 1. !"l Ijl CD !■- tv I. .'. in ■St ^r ro OJ -I— t CD i i '■■ TT ST Tt ST sf Tr TT sr ^ *j *\t ^r fvj cn Ul 0T cn Ui iji cn cn CT iT'i iXi iTi C"i iX IX X' IX iX X' iX CD CD 'X X' '3J GJ 'X CD X' Xi l—l X' Xi 'X l^J r : i^.! LU iX) l^_" CD iX IV 1 CD CD 'X' CO CD iX iX LU LU LU UJ l_U LU LU LU LU LU LU LU CM N- CM OT rv 'XI LO ro Q ST CM ro •■— * »— i ro "xj" m TT 'j.. OJ 0"> 0"i CM i.n CD LO 'jj ro i.n LD CD X' in 'Xi ro ts LO in LT LP TT ro CM ID IV ■^t ■— « IV CD ijj CO TT CM X' 1 1 : CO ro •^-i cn '•D rv CD KD CO CD •X' UO LO LO in TT Ti" lo UJ CD CD 1 1 1 CD UJ 1 1 1 UJ i ■ i ITi 1 1 1 Ui cn ij i Ui cn cn cn Ui a. iji cn cn iD CD X' iX CD Xi 'X' Xi 'X' X' 'X CD r- CM ix in ix lti *-< iX CO IV Lfi TT Q iTi ij. iji iXi X r- iS iX CD Xi iS iX in id IS >s CD LO iS X ffl Xi UO iX ro iX If 4 4 CD ro IS ID IS Uj iX 151 Li <£ CD U. X CO H 1- 2: : :j XI m us *. Q Qi ■I Pi LU L'J .>- LL! >— " I Q£ !- h Qi LU LU l-H 2^ u Qj £) UD si ro (M ■■— < iZEZ 1 O Ui cn 1T1 m (M IS 03 'XI ^r CM IS 03 UJ ro T-H en IV IV r-- I.J3 UJ UD UJ CO m in LCi in xr \r "*^ -< UJ E 03 U3 133 -I ro ro ro 1T1 iv i£i "^ — « CO i"J in UJ ro F-- r>- 1 1 IT\ 03 IJ3 1 \ 1 ro ro CM 03 CM ^ 10 ivi LP CM ^-1 iS cn r-- in ro CD CO LO CM cn in 123 iT'i 03 U3 LP \3 ro CM CD CTi 03 t.j-i in p ■1—1 ■»— 1 - — 1 ■^-t T— 1 •r-* >S G) Co ■s C 1 O'l 1 II cn 1X1 1T1 cn lj, cn U, cn in 0"i 1S1 iS iS 133 iS iS Gi iS iS C3 S •S iS ITl IS iS IS CD i : !-J IJjJ i"^ 1 CD GI iS G) Gi CD 'XI III UJ UJ UJ LU LU UJ UJ UJ LU Ixl Ixl in cn ^r ■X in IXI in CM in '•£) 03 ro Gi c-.| !\"1 1x1 '"'j Ul CM CD r>- ■jD ~r Ul 'S ■t Ul 1T1 rv Ul *tf '■t '£< CM ■^-i ro r-- in . ^r in >S 'T, IV in Oj CO "•\T ui ro «J3 cn *— 1 CM T— 1 T'- •^r *-i CO tr *-i IV if iS UJ ro IT, Ui 03 03 1 ! ' r- [•■- IV 'XI m in 'JO in *3 cn cn Ui Ul Ul Ul Uf cn Ul Ul cn cn CTi IS .::■ iS iS iS i7J G) G 1 iS C' iS iS '31' j ■■-> ro [V !?-! Ul h- U) r\- CS TT cn m 3 m iS in 1— 1 UD e-i 03 -r ^-* |V I'D ■-I 3 iTl r-- 11 1 'J OiJ ■<— 1 Ui 03 I'- in "5T ro 3 Ui U'l Ul Ul cn Ul 03 03 ll CD 03 CD CS >S iS <:: is is IS CS3 i^.' ■s in >s is Si G) m rn d CD iS iS LTi © ■si 152 in in in _1 LU H- UJ a d 1 1 Ql ■X 1 o U. ». 1 X 1 (0 f- i h- "7" | 1 O t Xj n i CO i UJ •. i JL Q£ 1 •X ! C-i LU 1 LU >■ i LU O > 0'. LU I — r i 0J X ' z: i U_ QL I en i LU _ I i — tT^i i — < l±j x z: i cn en o i i u LU a-: n_ CO 03 I LU ! I i\l i I LU ! LU I CO I i O ! _J i — * :T"l m i r~ CM r-- ro U'i iv M ro ^— i *-H ^— i *— < ! ^r iDD ^-» in on CM ID 'X 1 "TT CD CM 'X 1 in ■^t \r r-,-. '.■■j r\j — i ~-* i-- .1 U"i < ■ : 03 i ~\T ^r ■*J \r "^J" ■*t -nT '-; ro ■'■ ro i S3 CTi iS Si Sj iXi O is iH.' CD iS IS ! IJv_i CD Gj Gj CD GD m LjJ UJ m < - t LO |V :T. m XI ir. in CM CD i . n iv ■\T •■— r ! i - * ' CM r- rv_ rv- co iX' CO IX c_- iS iX' ■X 'X' 'X : CD iXi 1.1.1 LU LU LU UJ UJ LU LU III LH CD ij*i ■^H CM CD [V 'X CM t— J i_i i Q i in CM ~-i ro IV ro iX LO CM ^ ro CD CTl 'X o r ~ 'J ro ■■— • '■j'\ ! i_i ro : i ?•- 1" LO cm en 1 1 ! CM en ex 1. Si iTi -r-; *-H a C£l 'X en iTi ITi CO IV r-_ iv '■'- iv CD CD 'X '•j" ! 'X 1X1 'X c co 'X' iX' iXi iX 1\« iX' G3 IJ^' 'j^i 'X' ij^' Cj3 iX) iS iXi iX' :Xi CD iX' iX o 'S iS >i} iX' CO 'X' '-' s '-' hi UJ LU LU LU LU 1 J UJ ID ID LU LU U i m ID en ro CS3 iM ( 3 UD in ro cn •i—) IV Uj cn ro m en ■'— • ID iS CD in in Ln IV in IV i~i U'I r-- ■^r IV CO -tf !"*_ in 0j T; i— 1 ro rn T-l IV — ' CJ fH UJ a •■sT ro o Hi ro i-i en ID ^r •»-^ OD ■T *~t rv ro iTi in ^ ro »h IXi ■j) 03 "Xi IX' t p-.i >— ' IT. ■^-i -^H T-H — i *-) 'X i' fT; cs •X 'X' 'X cn en Ul IJi en en CTl ij'. U'i Ul G i en 0"! Qj iX' t±-l !_-.' is — . ro iv ix in ix in is en r-- in co en cn cn r-j CTi •^-1 'D CM CD ^r —i r-- ro 'X "sT V-J —i Ijl ITi [V L0 "3- <"'"| iT'i cn cn iX CD IJJ LU UJ UJ 1^.1 iS iS 'Xi ijj IS IS iXi iS Q IX' IX iS ce eo LO M-l L"i 'S in LO IS iS IS iX CD lo is in co 153 ,n a _j LU 1 — 1 Li- en h- n cn I - [ ij) cn lT| iTl cn cn IT, IT' ITl U\ O", cn cn U"i Ul CTi iTl iTl cn >ji Ul cn Ul !Tl ITl cn cn U'l cn Q~\ iTl U, cn cn cn cn Ul iS Ei CO KJ Ei Ei Ei cn CD CD E Ei '-■- ~ - 1- -l " LU LU IjJ Ijl |T'i cn 0. cn cn cn Ui iTi IT: cn U iXi Ij"', cn IT, Ul U'l cn cn iJl cn Ul cn 0*1 cn Ui Ui LU 23 Ei iS Ei EE E> (S 1!) LU IJJ LU IS Ui E Ei IS 'S IS UJ Ei E< Ei IS E Ei E' E' UJ £J Ei IS J IJJ ix! UJ LU UJ UJ UJ J E' Ef in en •a!' iTl 'X' CD J Ei E' CD ® ijJ Gj 'ZL J Ei 13 CD E lID G3 '53 J Ei '■~ '33 c:< IS i53 IS) j e »S i_D cs S <£ CD H ■" ' _ ■*""■ r-H ■r- 1 "~" * ■i— i E 'x' Ei 53 *S E' 53 E' IS Ei U i u > ■-i E E ro ^r IV ro CJ ro Ui LO in r-- i-* co IT' in r>- Ui 1— 1 "\J" IV 5 T cm S 1 : 1 ijj ""J ro »— i Ui o. Ul ITi U, CO CO on in CO tv N O ■S iS E E Ei E' IS CD cs iS iS Ei E in e 53 LO 53 in Tf LO in 154 CM CM Q S3 co 13 Si s iS s iS S CO iS S3 D co IS CO s s CD s iS iS co iS iS uu LU ijj ijj LU LU UJ LU LU UJ UJ U.J lj.. 1 i-Q CO CD r-- IV UD LD LI LO ■* ro ro CM IV ]V r- r>- IV |V IV IV h- IV r>_ IV IV cn iTi cn en cn cn 01 cn 1T1 cn cn cn cn en 0"i Cn U1 ■j~i cn cn IT. 1T1 1T1 01 cn cn en cn 1T1 en en en cn cn CTi cn cn Ui ui en en 01 en en cn en IT. 01 01 cn 01 ui iji l ; i en en en en en cn cn C, cn i Ul 01 en CTi IT- cn cn lT| cn cn cn Ui LLl ft Q LU M LU 'S S3 s 'S S3 CO iS CO is CO LU LU LU lu UJ LU LU LU LLl LU 'S eg iS 'S iS s CO CO S3 iS S3 'S iS iS S' co CO S' IS S3 'S CD CO 'S CO CO iS IS CO iS 'S S 'S 'S 'S •s CO 'S IS iS iS s iS iS SJ co iS IS CO CO s CO 'S s CO CO CO iS IS iS co CO S3 S3 CO 'S s CO CO 'Zj iS CO CO iS LU CO iS p _l S3 CD UJ o LU > LU •— i G/ £K LU I— UJ IS LU uj i_ i: tl a-: i K S3 SJ iS LD S3 S3 CO S3 LO is in IE IS S3 LO 155 CO Pi _J LU Ll 01 111 or iu Ct: u: IU f— lu m tx 3) CD 'S3 CD '3 CD 13 CD CD CD iS '3' 'S i t 1 '3 3 '3 '3' CD CO 3 CD CD CD iS '3 CD 1 o 1 LU UJ UJ IU IU LU IU LU IU IU IU IU IU ! '— < 1 LO ; T jr^-j 13 CO 'J3 iT cn IT iT IT IJI IT 0"'i on iTi cn i Cn iTi IT cn iJi iT IT en cn CTi cn Ul cn t is 1 I iTi cn IT CTi iT Ul en cn cn cn IT cn •: tj 1 Cn cn IT iT 'T cn cn cn u") cn iT cn IT 1 'M ix- '3 CD CD CD ■3 CD CD '3 CD co Si '3 i3 CD 3 CD 13 CD '3' i3 CO '3 i3 133 '3 '3' CD 13 i3 IS '3 '3 '3 2jT o LU LU IJJ LU IU LU IU LU IU UJ IU IU 1 1 j '— * iTl 03 i o 0~i IJi 13 CO CG 03 O.i 03 CO 03 ._l Ui 0"v IT CTi IT cn IT cn cn ITl CTi iT 1 1 1 _J iTl iTi iJi £ ?5 j-f J4 J4 J4 Ji IN* 54 "\.-:- en _i UJ I 1 U_ > UJ t . •— Cii / CiL Lj Ql 1— UJ UJ LU 1 — < 1UIZ > i if) o; Xi cn Xi cn cn cn i , i ui cn cn IT Xi cn Xi i Xi cn Xi cn ij , cn IT, X) Xi X, ui Xi Xi Xi IT, cn cn i n cn ITI cn Xi i r. Xi Xi X, Ui Xi Xi ij i ij i X, u i cn cn Ul ui U) Xi ' J ' CO IT. CO — 1 CVJ ro LO IV Xi r-l TT f- r is Ij ; 'X' \i CVJ CO UJ 'O Tj ro T-- Xi c CO CH Xi X, Xi Xi CD CD IS UJ CD iv r — CO co CO CO iS CO CO CO iS 'S iS c CO CO CO LO CO Ei iHi in CO CO ,s LD CO 'S CO iS CO IX, IS 13 in _J Ul IXj Ifi m U"'.i in m LT< in T xr xr XT 1 — 1 U. U) a-i 1X1 Ul Ul Ui Ui 1X1 i.'i 1X1 U, a - -! T~ 01 1X1 Ui cn Ul cn Ui 1.T1 Ui 1 Ui Ui Ui 1X1 1X1 1X1 cn CTi Ui Ui Ui Ul ij i Ij'l cn Ui O IX, Ui Ui 1X1 1X1 Ul Ui Ui 1X1 Ul Ui Ui Ui IM Ui 1X1 1X1 cn CTi Ui Ui Ui Ui cn Ui U) Ui CM 3 3 3 CD CD CD CO 13 3 3 3 i3 X' CD CD iX' CO S3 3 3 3 3 CO 3' 3 'X' CD 3 rx X: Xi 3 3 3 3' 21 O UJ LU UJ UJ UJ UJ Ul UJ UJ Ul UJ UJ Ul 1 — 1 CM CM CM CM CM CM 1—1 1— ■ ■•— * H *— I 3 3 _4 pi 1X1 Ul cn IX) m cn Ui Ui Ul an Ui Ui _l Gi Ul 1X1 cn 1X1 cn Ui Ui Ui Ui Ul cn Xi ' — 1 1X1 1X1 1X1 cn Ui IX, iX Ui '1 Ui Ui cn 1X1 Ul T ' Ul cn iX. cn m a, Ui IT, Ui Ui Ui Ui Ui l\l Ul cn cn cn Ul Ul Ui Ui Ui Ul U; Ul Ui I—* "X 1T1 cn a-i cn Ul cn Ul Ui Ui Ui u Ui Ui 01 1X1 1X1 cn 1X1 Ui Ul Ui Ui Ui Ui 1X1 U1 Ui UJ 1^*' 3 CD CD CD CD CD X' 3 3 31 3 3 _l M— < u. 3 CD CD CD O CD 3 'X' 3 3 3 3 3 © CD Si CD 3 CD iD CD 'X' 3 3 3 3 21 UJ IJj UJ UJ UJ UJ UJ UJ UJ UJ UJ UJ Ul i_i CD CD CD OP CD in CD 'Xi CD 00 OD 3 CD v— < 1X1 1X1 en en Ui Ui Ui Ui U) Ui Ul Ui cn _J en IT, cn cn Ui Ul Ui Ui Ui IJl Ul Ui Ui _J 1 Ul cn Xi cn Ui Ui Ul ■X Ui Ui Ul Ui Xi •— 1 cn cn 1X1 iX Ul Ul Ul Ui cn Ul Ul Ul Ui XI cn Ui CXi cn U) Ui Ui Ui xi Ul Ui Ui Ui Ul 1X1 cn ov Ut Ul Ul Hi Ui Ui Ui Ui Ui — ' Ui Ul cn CTi Ul Ul Ul Ui Ul Ul Ui Ui Ui 'X CO © CD - CD X.i 3 X' 3 3 3 a-: h- i-j ■^-* '■D ro •T N ro CM M •3 m LO r^- a: 'X CD i33 i-H IM PO in r- Ui *~l xr i'- 3 U_ CD TO ID 1 CM 'X ■Xi 'Xi ~l ro —1 it, 3 '3D iX 1X1 1X1 Ul Ui CO CO ■T: CD CD r-w IV ._! _J t-* 3 iD CD a CD CD 'X' 3 3 CD X' 3 •X u UJ n: Ul t- ■X J>" >^ ?--- >■* n K «v ^'. *t ?.; }.:> CK a 3 3 (S3 x. CD ID iX 3 3 3 CD CD 3 in 'X' in CD ld 'X 1 IX' 3 in 3 U'i X Or. O 3 3 t-l *-< CM CM ro ro Xf ■tf Ul in ID IX IX Ul 158 co e> UJ Ll _ ^ Qi <£ i CD r— -o »o kO LU _1 CJ LU CO I _-! I S3 I CM ! CM i 3: I D LU I IM i LU I CD 111 a QL Q a-' a: til LU |\- S3 I'D CD 03 CD CD E. iTi i -.i iS CD iS Ei S3 CD CD CD IS IS! 'S f*D LU LU LU LU LU LU LU LLi LU LU in ro *-i 03 m »— i iS ro CTl xr CD iTi co 'J3 LQ rx CM *— i cn l"Q \j po ro ro ro IO ro ro CM 0J IT DO Ul U i 00 Ui CTi CTi iT) cn Ul 00 iJl cn Ul ITI 03 CTi iTi IT, !Tl 00 cn cn 00 Ti 03 Ti 00 m iTs iX, ' Ui cn ITi 00 t i'i IT) 'Ti en iTi Ul Ui CT) CTi Ul 'Ti 00 cn iS CD ijj iS iS iS IS iS CD tS O CD CD iS .i 3-': j-i ;•-:> ?i >•? i^ S CD iS CD m iS (S3 Gj 'S CD CD LO CD in CD in CD IT) iS lt. CD CD CD -> -i CM CM ro ro \t ■t in CD LU LU HO ro kD in f\| CM CT, CO CT' 0"i cn 03 IT) cn iS CD 'S CD CD 'S iS CD CD CD CD CD c. CD iS CD CD iS iS iS CD iS CD CD iS cs , 1 1 Lu LU LU LU LU LU LU LU LU LU UJ u cn iTi iTi cn cn CD CO oo no |\- r>- [\_ r- IT IS CD CD CD CD 00 CD CD ao 03 CO o iTi cn cn in cn CTl cn cn 'Ti cn i iTi IT m iT cn IT) CTi cn iTi CTl cn 'Ti CTi IT, cr oo U) cn CD in C) cn 0") IT, CTi Ti en IJ in >n cn cn cn cn n CT] Ul cn CTi Ti IT iT, 01 cn iT, 'T) cn Ui CO Ti cn cn cn IT iT. IT- 'Ti cn U. cn Ti ■n cn 00 ti i iTi CT iS CD CD CTi CD l_v_l CD a CD iS iS CD IS iS CD iS CD CD CD s CD CD CD CD CD CD iS CD iS iS UJ r-- IT-l UJ LU IV cn LU r-- 00 LU LU LU Ul UJ UJ IN- CH LU LU LLI in 00 00 00 cn 00 cn U ! 00 ui in 00 iT) cn 00 CO U ! 00 1 1 1 00 cn cn m m m m iji cn IT, cn CO 00 cn Ui ui no cn cn 00 cn cn CTi cn cn 00 00 Ui CO iTi cn cn cn cn cn 00 m 00 cn cn cn CTi m CO m 'Ti m m 1T| O) CTi IT) CTi Ul CTl U 1 cn m cn m CTi cn cn lT| Ul cn 00 CTi IP »- i-£i m TT r>- ro OJ ro OD in in r- CD iS CD ■^i CM 10 LO t- Ui 1-4 xr r-- r"i iS ■Tl CO ^T <:■•! 'S Qj 'JJ -T ro T~, cn no INj.1 Ul Ui cn CO *Xi CD iX) CD iX' Ei CD CO CD 'Xi CD CO El CO "X CO 'X' CD iX CD 'Xi 'O LU LU LU LU UJ IX Ld UJ UI UJ LU LU cn cn cn CD 1 1 1 r-.- rv. 'Xi ■»Xl IX' LD xr |\- r-- i\- |\- r- r-- i ^~ |\- r<- r>~ IV IV. cn iXi •Xi 1X1 Ui iTi CTi tXi cn CTi iXi cn cn cn cn cn Co 1X1 CTi cn IT. cn cn cn iXi : cn Ui cn UI Xi cn ,T-| cn ui cn cn CTi On CTi cn CTi i cn n-i IX. cn o-i cn cn IT; cn 'Xi iXi a i cn G i 1X1 cn iXi cn a •Xi cn cn CTi cn cn CTi cn cn CD CD CD CO CD CO CD >\> TO «D CO iX' iXi CD CO CD CD 'X' CD IXi CO CD CD CO CD CO CO ui LU LU LU LU UJ UJ LU UJ UI UI LU LU LO LO LO LO LO LO "xT XT ^ V "\T "3- -tf CTi ITi cn cn cn |J; u i cn cn cn iXi lT| cn iTi CTi cn cn cn iTi cn iji cn cn iXi Iji 1 1 1 cn cn cn cn m cn iT. CTi en cn en cn cn cn cn cn cn cn ui tn CTi ij ) en iTi ITi cn m cn cn cn cn IT, cn cn cn Co CTi iTi cn ITi cn 0"i cn iXi Ui cn ITi ■ ' i ij; cn cn cn IT"! cn cn cn ui iXi cn cn (Xi cn cn IJ'! iX; CO 'X CO CD — 'Xi - f CM ro •X) E' CD « OJ ro LD IV. cn ■^ •■-a- E CO LO ^r C-l Ei iX 'X' ■\T PO ■^H CD cn cn di IJ 1 cn CO CO co CO CO Ei CO Ei Ei }•? *S >> J4 a 3vt X >* E CD CO CD E CO E E CO in CO LO CO iXJ CO IT) M 160 CO Ei CO LO IX - ! CT) U. ■en o en 00 UJ Ui CI LU ^> UJ i— Q£ Qi LLi I — UJ QJ inn z: ^ LL O I- h- =s LU LU LU 0" ZT LU UJ I— O LL Lkl CTl CTl CTi CTl CTi CTl CTi CTi CTi CTl '. "\l 1 S3 CD CD CD CD CD 'S CD (Si Ei Ei S3 CD CD iS CD iS IS CD S3 Ei Ei EJ Ei [ CD CD IS iS iS IS IS IS 1S1 E' E> IS E : O I LU UJ UJ LU LU LU LU U UJ UJ UJ LU LU t— * 1 ro CM CO r- LO ro ,— ; s 0j '£i \r CM rvj C\i CM ■*— 1 i—i T— 1 — ; r— i ■^-1 E 1 iTi |T) f\"l 1 | CTl CTl Ui CTi CTl CTi CT'i CTi IT, CTi CTi CTi CTi '— ' 1 CT- CTi CTi CTi CTi CTi CTi CTi CTi CTi CT, CTi CTl 1 LU .■• 1 CTl CX': U'i CTi CT, CTi ui CTl CTi -CTl CTi CTl CTi 1 r j G 1 CTl UI IT; CTl CTl CTl CTl CTl CT, CTl CTl CTi i ' — ' T i U, CT, CTi CTi CTi CT, IT, CTi CTi iA CTi CTi CTl i ID CTl 1CT1 LTl IT, CT, CTi CT, CTl CTi CTi CTl ij'i CTi i LU CD CD CD CD CD iS CD CD Ei CD CD i~i CD 1 ■-"* 1 Ll O CD CD ■S CD CD CD CD CD Ei iS CD - f~- r-- r- IV IV ._l 1 1T1 Ui ui CTl Ul CT, en CTl lT, CTl IT, 1T1 CTl _J 1 CTi en CTi CTi CTl CTl CTl CTl CTl CTi CTl CTi CTi ' — ' 1 CTi Ui Ui CTi CTl CTi CTi CT, CTi CTi CTl U 1 U ! Z_" 1 (Xl CTi Ui CD CTl CTl CT, CTl CTi CTi CTi CTl CTi IT, CTi CTi CT, CTi CTl CTl CTi CTl CTi CTl CTl CTl —> 1 CTi CT; CTi ui CTi m CTi CTi CTi CTi Ui CTi CTi CD 'S CD iS iS 'S' 'S O CD i Ql 1 a 1 l- 1 1' 1 CD *-H '.£1 ro ^ rv ro CM ro 00 LD in IV 1 a: CD in IS 1-H CVI r-D ^ C'J IS CO <£j 1- ro -1 CTi 00 CD Ui CTl CTi CTi CTl 00 CD 00 GO 00 h- r- 1 _i ! _J ■>— * CD CD >S iS CO IS £3 Ei iS El Ei 'CS 1 a 1 1I1 1 QC 1 LU 1 1- 1 ? 4?. J^ i Q£ 'S CD iS CD CD CD CD CD Ei iS E E E^ O in iS LD i33 in IS IT 'S LTi Ei ID Et 1 Hi • ■ • - • 1 CD CD ~- < t— 1 csi C'J ro ro ■^ •i in LO i.r; 1 QC 1 QC 1 LU 161 Lt U" O UJ t— I W UJ ft ui Ctl UJ (— LU CD U.i LUlI CK 'I _3 ■ 1 Ll D£ o h- 1- "_j UJ IU UJ CO 2T L'J UJ 1- O ■Ot Q£ JL ;Cj — 1- 1- 1- y" 03 03 03 CO ■X i — i t_i ui Dx a i — f0 an CD CO iS 'xi 'S IS CD 'S ij,j '3j LiJ Ci Li.1 i a i LU LLJ LU UJ txl UJ UJ LU i-Ll Ui Ul LU UJ i '— * i i Ch IjO cn iTi T CO ro \' s - IJO 'S Ui 'or i _i i 13 kfj CM f-w OJ l\- T-l cn CO •r--i '"!' LO r"~ ) j | ^ J Ui CD CO LO PO CM CD i~Tj r\_ LO ro ■^K • i—i | ro CM O,! •'■■1 CM l?-J CM CM •i — i i-< -— , ^— ! ! H : cn Q~i cn CT, ITi c . ITi cn cn cn cn Ui Ui Ui Ul Ul Ui iTi IJ'I cn IT, iTi Ui Ui cn 'Ti i G3 1 ui C r i ■_' 1 cn s CD 'S iS CD CD iS g;i s CD ! ill' 1 UJ LU LU UJ UJ UJ Ul LU UJ LU uj UJ Ui i '--■ 1 ID CI CD IV LO CM U'i CO ro iS i\- ro CD ! _J ! f- P-- p,_ CO !,_£ i.ri LO LO 1X3 LO '\T -sT \i" ! i 1 cc 03 '3-i CD CO CO G3 CD IS CD 03 03 CO 1 l ™ 3 ! CD £7". Ui Ui JT; cn cn U'i Ui Ui Ui IT, cn I *Z ) Ui ij'i ITi CTi cn -j! Ui Ui cn Ul Ui Ui Ui cn U'i cn iT, iT, cn cn Ui Ui Ui cn U; Ui 1 "^*" i cn Ui 'T*i IT- iTi cn IT) en en Ui Ui U i cn Ul Ui Ui IT: iTi Ui iT; cn Ui Ul Ul cn u CD ED IS IS cn LU a: a lu ^. LU — u. Cn LU k- uj in lu T. T. a. IX' G2 ' CD -*■' — ' — ' — ' -"■ — ' -*■ ^- -* ' " u ' II i LU LU UJ LU UJ LU UJ LU LU UJ LU ! CD CD rv U.I ro ^-i rn LD «— i IT, T cn i in :■<'• (M *— : CD cn rv '£' in ro rvi m i Tt ■yl Tr ~r -T ro ro ro ro ro ro ro i ITi cn cn iTi cn cn lT| lT| cn cr, cn U 'I : cn iTi cn iTi en cn cn iTi cn iT> U'i IT'I | cn IT! cn ITi cn cn l^ i iTi "T'l f\ 0'- iTi j cn IJ : L : 1 cn cn cn Ul cn cn cn 'i IJ'I ! cn iT; IJ'I en cn cn cn i cn IT'I CTp CTi i I 1 CD CD CO CD CD iVl CD CD i_J i>i-i Gi G i CD 1 i CD CD £T X' CD CD CD CD i\ i iii CO CD ■3D 1 T. I LU LU u LU UJ LU LU UJ LU LU I.U UJ U i ! O 1 [V rf T-i CD LD CV1 U 1 i.£i 10 CD ID ro Ul 1 — 1 VI? '.ft 'D LP LTj IT) "T -r \t T ro ro C'J 1 ._) ! CD CO CD CO 1 , : CD UJ CD \T>. ITi CO CD CD 1 _J 1 iTi iTi cn iT', cn cn ui en cn cn iTj 'Ti cn |- —, | cn cn ij> cn cn in CTi cn cn cn iTi IT. 'X i E i IT'I cn cn cn Ui iTi en U'l ■Ti IT'I cn CTi ■Ti I I cn cn cn cn 0"i iTi cn m cn IT, ■ 1 1 cn cn i — i cn cn an cn cn cn cn cn cn cn u'i on iTi i i CO o Gj CD •S3 i3 CD CD CD G- Gj 1X1 CD CD '•D ro ^r fw ro r i fO 1X1 in IP r- CD iS Q ••■< CM f, LD f'*- Ti <-» ^r r- m CD CD >-0 \r r*j IS CD '■D ■tf ro — i iTl on Gi cn iTi cn cn iT. UJ CD 03 CD CD IV IV UJ H CK O Q£ Ql LU X K >« K' \-i 5>* ■x w CD ISJ O IS iS ISJ CO IS CD CD LTi CD LO s in a LO 'S (E w ro CS CD LO in 163 CO n . i LU *~4 u_ ■%- _ LU 1— < Qi X K LU 1- H lu m 111 !_u LU jZ t __ i— r Ct ■i lU LU LU rn y 1! 1 IJJ Ix! i i !n a-: |V ! — :Q ■-> LU i- i — •i ;r CJJ u.i Ol t- j Lj LU Dl a to o h- yj Hi a U a U- o O LU . r G£ a ijj IS' CD '-'-' "*■ ~^ *■"■ _..> ■*■ __ -"•' -■- -*- 11 LU LU Ijj l_U LU Ijj LU LU LU UJ 111 rtr, CD CD )_!_. © CD I'-. I CD G? (3 63 CD 'Hi GD CD G3 CsD CD CD GD CD I'D CD CD :"'"■ 'ID O iS I_i-' CD CD GD CD CD CD CD CD iX" :!D £) '3 ■— i T- i T-l -r-! ^-: *«i ~ 1 ''. "T t-: *^" CD CD CD CD '3 LU CO CO in CO i2 LU l_U ^, j= S3 CO I^'i CD CD CO 'X' LU LU CD CD CD CD CD CD CO CDi CD III LU hi CD CO CD n LU IJJ ( 3Z' l^J 1 CD CD CD i~"t S3 CD LJ IS CD CD CD UJ CD CD CD CD CD CO —, LO *-> IV m (VI r\_ in OJ 'S C; Ui en en CTi CD CO IS CD CO in 164 LU CD CD CD UJ LU IS CD IS CO CD iS ro n m m r-- i\- * V jj >s ;-t >i }: > <_ 54 >£ D iS iS is CD CD IS CO CD IS i^-i j in o in iTl in CD m CO in IS CD in LU Cl o li- en 3] r LO LU * £K Ck -•— LU >— « QiX Di LU i— h- LU CD £V LU IjJ L E" i_ QL i UC iV 1— (-1 •— ii l- 1- ■! 1^3 w CD CD CD CD LU IS CD CD LU UJ CD CD UJ LU l-LJ CD S3 CD CD LU LU CD 133 CO LU LU IJvJ CO CO ijv.i i\j CD ix.i !X.i LU CD (3 CD CD CD CD CD CO '3 CO CD S3 iv, is CD CD CD i3Z' iD CO CM CD CM CTi © in *- r-- m CD pi- LO CM CD CD u, CTi iji iTi ;~'j LD in m CD on IT; CD CD CD CD iS CD CD CD CD CD LO iS iS co is in CO LO *-< ^ P.i -J Tf 165 LU CD CO LD u 'i ro ro n r- in r - r- rr 'S LP CD IT- LO LO LU <:i er > LO i— 1- r I i i ■:j £ 01 LU * (X ,_L. ■X i-i LU LU 5— 5 •._• LU >-H Qi X U: lu h- |_ in ca ct ixi LU i V" , , T~ D_ a: X LU 'X j.- X X h U_ IX D 1- !- O :z.' LU IJJ 0') x LU liJ LU O X a Q£ !- O — LU 1 XX LO iT) 01 03 O •! 1—4 u Ui 0* Q_ u O QJ <0 m Lu LU •x Li_ XT O I- CJ x u. .-J _J a u LU IX UJ I-- ■X ll: Ci' a LL liJ l_Ll El iX> IS '3D a CD E CD CD CD UJ UJ LU UJ LU UJ UI CD CD O E CD CD E 'X E CD CD CD Ei E E E '31' CD CO (53 Ei E CD CD CD CD Ei CD E CD 13? E CD CD CD E CD ( xr E ■ ;_j CD E E E E CO CD C LU UJ LU LU UJ LU LU LU LU UJ L 133 (S3 E (S3 CD Ei Ei CD CD i-i- ■ C E' E E E CO E E CD E' E C E CD E E E E CO CO CD E H CD E E E CO CD E CD |V| CD i i"j i- -i E CO CO CD CO E r~ CO C E [ E CD CO E E CD E E CD C ill E CD CD CO E IS E E E L E E E E E CO iD iM E M en ■—< IV 03 CO LO — i r-_ ro ■i— > 03 UO (S3 IN- in CM E CD LT" n (S3 CH 'X. >T> '7, X: Xi 03 — E CO E E E E CO E E E E E »- (M CM 166 E E E E' E o (3D <3Z' Ei E E' E GD I^J G3 LU LU LU IJJ UJ LU LU UJ UJ LU UJ LU IJJ E E ■E 'X' ■£i iX 1 CD E CD E K' K GD E E E 6D O E ■E E E' E' iZ< >Sli (52 3 IS E iS 63 E ■-' E CD E 1X1 GJ '3Z- E E E j_D (3D E CO E E E G3 i3D t3Z' Ei E X> , .' J . 1 £~i E CD E E CD "ID ill' ■ ! j, r E E E S3 © E ■X' i_J E' E © iZJ iD CO E Xi S3 CD E CO E CO IX' G3 GJ '2l! *""* 1 ' ' ' -— H ,— < ■'~ : ■*■"• T ~ l " ,— ' "~ 4 T— J T-~ • *"" ' M 03 >jj 1T1 "\J" ro ro ro cn n. LO ro r»- IV r-- r-- CO K J4 J« « X « N 5': VJ >S }••■ ?« x 'X CD E E CO CD eo E El E IS E CD G3 IX (S3 LP CO m E in CO LD S in IX' en a UJ o U- 1.0 > w LU i — • ix x Ct' LU 1- h- LU LU Ql LU LU i_ 11 ■— ■ i_ Ct x x< m x XX X U. LL IX a f- 1- □ X UJ LU CO x , LU LU LU o X — ■ LU ££££ CO LO LO CO O X *— t O LU Di a. -o o jo LU LU Q£ O h- CJ a LL a u LU x LU x Ql LK a Ck LU LU CD S3 X CD CD — iS UJ CD LU IX. LU CD CD CD CD IX IS IS CD iS >S CD 'S CD CD CD CD X' CD LU LU X' CD CD CD LU UJ CD CD IS CD iS X 1 IS CD Xi CD CD 'S LU iS CD iS '-^i •3D ■ID ijD O iX: UJ LU LU UJ lli GJ ss CD G) CD i3~' iTl iV: iVi IVj G3 © CD CE iTi CD ** ■* 1 "~'~ IJJ cd iS X 1 vl X' OJ ui ■»* r- CO X 'Si *-H P-w TO th 0J 'J3 iS IV LTi IM IS CO LO ro lit iTi cn cn cn 03 0j CO CD X is >z J.= ii J;: }■': iS - ■T- h- ■31 (I O !_■' Ct- a o LL ». CO i- 1 2: 1 n :.:> 5J ...i-i LU ». ii Qi UJ £ > w LU *— ■ CK 1 Qi LLI H h- UJ 13 Q£ UJ LU x_ £_ >— • .£_ Lt ■-.:- ._' LU 'X £ .,.1 u LU Qi □ h- !- i I : m IJJ CO y~~ LU 1U IJJ O X a; o. t — £} -< LU h- ■X "C LT' en co CD <-A 'U_ t — < LU a lL aj o CD 'jD h- u _J Of _J l_U £j G3 IS) I'D |T| 13 CD 13 i3 13 13 '3 El 13 **-' - 1 - - 1 - - 1 - — ' '_*.' - 1 -' -*■ *-*- -*-' — LU LU LU LU LU LU LU LU LU UJ UJ UJ Hi a Ul IT; IT'; 0", IT, U, i Ui I 11 Ul Ul U'i m U'l Ul IT-, U'i U'i Ul U'i Ul Ul Ui u IT, iTl U '1 CT'i U'l Ul Ul Ul < IJl Ul Ul Ui U'l Ul 1 I 1 U'l Ul o-i i ii Ul Ul Ui Ui Ul U; U'l CTi IT, CTl U'l Ul Ul Ul Ul u. Ul Ul U'i i-Tl en U'l Ul IT, Ul IT, Ul U'i en Ul Ul Ul Ul iTi IJ'I CT'i IT, Ul . Ul U'l 1 Ul Ul Ul Ul IT 0"1 0"'i U '1 IT, Ul U'l U'i Ul Ul 1X1 IJ 1 Ul Ul UJ il.i 13 UJ LU 111 Cj iTi f ". 1 iVj : v . ; G l r j ) ... iS S3 i....i C5D CD Ul LU LU UJ LU r~i r- 1 |T"; '3 iVt 13 IT, 1 -. i 1 -. i 3 'X* "X t3Z> 13 13 L.' ^ ,~ ._ »-i 13 CM i3 3 rC LH 1T1 ,-, h- IO *-* uj 13 1 : 1 i,"i Ul CD OZi >X' '3D 'X' >S GJ 168 UJ LU 1X1 S3 'X' 1X1 'X' •X' Gj CD 1X1 iXj 1V1 -.— e en r~< S3 iX' CD CD 'X 1X1 'X' UJ UJ t Ll.' ■* 03 'Xi ro cn |T| CD 1J1 IV r-- ro iv LTJ 3 IS 13 in IS IE 13 LD 13 13 CD Q lh 13 iS 13 ITi i3 CD l~/ : ,", ~* CM CM ru ro \J \i" LTi LD '.£i 01 c _! LU 1 i U. ■v cn LU i Qi 0£ LU i— * Q£ X Cn Ll ! — f- LU LU cn CTi cn U'i cn IT'I 0"! cn i r, U'i Ui U1 Ui CD a-' IT, U'l U U'i iji C '; U'l U, U; it. U'i LU UJ LU UJ UJ UJ iT'i iTi IT'i cn Ui IT iT'j iTi u'i U'i Ui cn U'i CTi U'i cn U'I Ui IT) CO U'I cn Ui Ui CTi IT, U'I lT| Ui cn lT| CTl i U'i IT, cn IT-, CTi IX, cn U'i U't IT'i IT'! U '! u'i U'i U'i LU LU LU LU LU LU IS CO CO CD CO IS CO CO CD CO iTi IS CD iTi CO CD CO IS ITi to CD CO IS iS CO IT! CO l\ ! IS iS CO iTi CO CO IS IS iS co iS S iS iS LU iS IS CD CO s IS UJ UJ UJ Ui U..I CO IS OJ LsJ CO IS '3D (S3 S3 GJ G 'Hi t3J £3 GD Ei G3 Gj G3 G3 LU LU LU LU LU LU LU LU LU UJ LU UJ III iXf i2 GIf CD G3 G3 CD '3 iVt ® © © >_T' G3 Q £53 iTi GJ GD E G3 '3D J_D '21' G3 lV[ Q iS S3 iS G!l "J"' i3D G3 IS G3 iVt G3 )3~ : i3D I"Tl GD iS S3 GJ (T"! G3 l"^~i GD tVl GD GJ 'S3 CD GJ (53 G5 i3Z' © CD £3 'X' (S3 GJ ® (S3 (53 iS GIi *—' T 1 ' ' "*"' 1 t ■r—J " r_ " 1 ■t— i -i-H •^H T- H ■r— * ' r ~ i CO CO iS CO rj CO CM U'i ■*-< IS in ^H r^- ro T-H CO r^- LO CM CO CD iS ui U'i Ui Ui CD TT T] ^3" UJ LD ro Ui IV. in IS. IV- IV — CO 2_ CO UJ «. LtL IX 'I pi UJ UJ y- _> "w UJ 1 ! u. j_ D_ ijj i— 1- UJ £ Q oc .y cm cc ~i cq Tl •_n iTl iTl Ul Ul Ul en iT'i Ul Si 1 1 Ul CT' ■J'l U"i u i Ul CTi Ul Ui Ui cn Ul C'-J I Ul U'1 Ijl CH Ul Ul Ul Ui en Ul !_• 1 lj"'i Ui '. '--i i " ,_" ^_* " Q ^ ,_" ^ ,_" V-" ■t—! CD 3 'Hi CO CD i~i '3D IS 'Si IS 'jD 133 3 o i LU UJ Ixl UJ UJ |J_1 UJ UJ UJ UJ UJ UJ UJ i i l3~l Q IS <3Z« '3 IS | ! | I 1 * - } l~] CO '3D |T") i?Z pr"; jT"[ 'Si f~ iS r--~: iS UJ Q M •3D is 3 Ij-J 'S iS iS iX' 3 T 1 CD GJ S3 CD '31' 'jZ' CD i'Si £5 'Si iS CO —' t— i ■*— ' •'-' —I ■ r_l ■^" : ,_ ' — ' ■,--: ■.— ] ~ H ■i — ! UJ _J 3 '3D IS: 3 IjD ijD '3Z' Cv 1 IS iS © IS Ijj U_ i3j -*■ =J>_ -i. -*- "'• ~ lJ -^ ■ ■^ Jl. -^ ~~ *7~ ( LU l_Li UJ UJ LU IjJ LU LU UJ UJ UJ LU UJ CO 1 — ' i 3 CD i2 CD C£ 133 IS 'S iSi iS CD '3J 3' | i rg i'Sj i30 U' cs CO 'S j J IIP 3' © i3D i3Z' iS *St IV' Cv' 3 ■— ' ! S3 CQ © i3D IS CO 'Si iS '31- '3 3i T~ j 3 CO '33' '3D iX» '3D •Si CTi 'Si iS i3Z" iS 3i — ' 1 '-' -r-H T-J - — ' ™ '•— ' 1-5 *"* f-' ■^1 -^-i "-' ~-. — ' G3 CD CD i3Z' i3D iS 1A. 'Si ~ 'S ttL o h- u CD 0J CO 0J Ul 1-i IV CO "d" Tj- 00 'Xi Ul ■X CD LD T^-l IV rri ■'-i Q3 i,n LT.i "*^j" ro ro ro u_ O r^- in PJ UIi Q~i in Ki "*-! if) r- m ro O IX! iTi Ul Ui Qj CO 03 00 rv. r- r\- r-- -J d _l ■>-- * CO E' IS GJ CO *s IS iS 'Z' GD iS 3* *: j£ ^ ; . 5^ 3-t 3* 2i'. "¥- ?v" j-* 5-£ a: CD 'S3 CD 3 GD iJO iS' CCi IS 'ID © IS 3 CD LP CD Ul Cxi \T< iSi LT' iS Ii"; O LT.i 3' ZiL » a © S3 n- 1 ..~) CV 0-1 f1 ~l m ■Nf '"d" JJ"; UO 'Xi ii: 0i LU 170 CO LU ft o u_ i-O I- cn UJ LU > ft ft z: cn cn iTi 0*1 cn o cn cn 'Ti IT', cn lT| cn 'Ti on 1 ! < Ti'! cn cn cn ON ITl iTl iTi cn cn cn ON cn cn cn cn cn iTl cn iTl cn iTi cn cn cn cn Ei Ei Ei Ei CD CD III UJ UJ UJ UJ UJ LU UJ UJ UJ UI UJ UJ en Ei CD Iji.l CD Ei CSD © *S.' CD 133 CD i$"' r-i a CD Ei CO CD '3Z 1 CD cc 'IT* GD G3 IS CD CD CD co CD Gj 'ZZi C£: CTj G3 iS 3 G3 Ei CD CD CD Ei c CD C3 GJ S3 GO CD G/ ro Ei Ei i3 CD CO Q CD >!D CD CSD CD GD CD Ei Ei Ei CD CD GD iV- 12 CD '2D KiJ 'I! 1 'S3 CO E' CO CD © '31' CD '2!' GJ CD G3 is;; '" " —l vH "■"' 1—1 —i 1 — i •i—i •,— i .,— < .^ *~ ' ■r-H El E CD © IS Ei E CD Ei III UI LU UJ UJ LU LU LU UJ Ei Ei CD CD Ei CO CD E Ej CD E' CD CD CD CD CD CD CO iT) Ei CD El CD CD CO Ei Ei CO CD CD CD CO CO CO CD CO CD E CD Ei CO to CO CD CO Ei Ei CD CD CO Ei CO iTi E El CD Ej Ei CO Ei Ei CD CD u UJ iVi lT"i G3 CD CD Gj CS3 ED CD Ei CM CO CM iTl — i r*- 00 T| r;- 03 i.£i i~i E LT< *— * IV ro —i 03 'Xi ijj \l ro !\' ro Ei r*- in CM E CD n ro -.— ON IV in PO CD cn cn Ti cn 03 03 CO 03 IV IV '■- IV —i CD CD CO CD E CO UJ CD CO If ' CO — -i CJ CD E Ei LT.i Ei LT.i E CO co in AT Ln 171 en ■ -.. _J LU 1 ! U_ ■ _ ■w LU r— i ■V T CK UJh- 1- uj m £H LU lU jr t~ >— i T" £V ' Ql Qi ! — £3 — UJ H ■" "Z! tn cr tn CD ft 03 I IJJ ITi LU LU _J a LLi LLi I- •X a a a: IX UJ IS LU © IS IS Lvl |— I GO £3 o iS IS IS iS Q IS -*■' "■*■ — ■^ -*- -*- -"-' — - 1 - 1 — LU UJ LU LU LU LU LU LU LLI LU UJ LU ■S iVi IS ITi Ui cn IT, CD i i i UJ rv. r-- CTi Cn ITi CD CD 03 CD CO on 03 rri 03 iTi l"T-< Ui iJi IT, cn '; |J: cn CTi 0*"' in I'm iT' cn cn iTi cn CTl cn cn Hi m IT C~l iTi Ul 'Ti Ul cn CO cn cn Cl 0-i iT, iTi Ll \ O'i IT'! I" 1 "'! cn cn cn cn iTi cn U'l |J cn cn 01 ij'l Ui cn ITS cn i ' 1 cn 0*i iTi 0"! 1 . 1 0"*! iT, Ul ui i cn cn ,; : Ul u_ iS >S CD iS S 'j IS IS E IS GJ Ei •S iS IS iS GJ 'S IS (S3 >S S iS GJ 'S IS CS 'S iS 'S iS iS iS 1—1 CM i p CJ iTi ■^ r^- 03 •M- IE LD ~* r^- ro ~-i CG U3 in IS r«_ in CVJ CD \~Ct LH ro *H ea Ti iTi ITi CTi t_:j UJ CO 03 -4 'S iS IS iT IS iS — , G3 03 ro r-- 'D ro IV 1 : > ro ro iS IS CS3 ill. CD 63 IS CO IS U3 CD iTi 'Hi 5>:> IS U3 IS IS in IS L?i IS Lfi IS IS CM ro r o Tt \T U""i L! _ t UD 172 U.l UJ en a Ll Z3 > i tj) LU * I q; lii I ■I I Q U 1 1 LU 5- i -•v. ■w ] LU 1— ! Qi 31 1 lii LU 1- i H u l CO Li- Ui 1 LU -■' > I'M ?' Ul UJ LU UJ LU Cn CO IV in Tj" !"-.< CM ^-i u ■ CO IV LO LD LO m in m U~ Ll" \r \i" Ttf- cn G> cn CH CT| cn cn cn cn iT ITi sT, :T, iT cn 1 1 •( cn i cn i < : IT'. CTi Cm Ul C i i i C~'i cn cn cn cn Ui cn cn CT cn < n cn ,Tl cn cn cn CPi cn cn i"n Ul CTi cn cn cn cn cn cn cn cn cn cn ; cn cn 'Ti cn cn cn Oi S3 Si S3 S3 S3 S3 S3 S3 S3 CD S3 S3 S3 S3 S3 S3 S3 S3 S3 Si S3 CD S3 1 : | LU LU LU LU UJ UJ ! I 1 Ul LU LU CI i cn cn IT Oi 01 cn cn cn cn cn cn IT' L'1 iT G ! CTi cr> cn cn cn cn cn cn cn G i IT, |T, iT cn cn iTi cn CTi cn cn cn IT; cn cn cn cn cn cn cn cn cn cn Ui cn cn U i cn ui cn cn iT m cn IJ ! IT, cn cn cn ,T, cn cn iT cn iT cn cn cn cn cn IT cn cn cn cn cn cn cn G i cn U '1 iT cn S3 S3 S3 S3 G3 S3 S' S3 S3 iD Si CM ID Si Lf3 ~ Si V in IV, Lit , j ■■ S3 S3 SS iD S3 J) CD S3 iD Si ~ IT) *4 iD in Si 173 'jj LlJ !jj LU iT t-l CD S3 S3 S3 ID S3 S3 S3 I^L.1 LU U, Ul LU UJ LU LU LU LU Ul 'D S3 UJ S3 S3 S3 iD S3 UJ S3 S3 'D .Vl S3 S3 Ki CD iD iDi Si S3 S3 S3 ID S3 S3 S3 ,Vj S3 .,' ■VI SJ S3 S3 S3 Si S3 CD Si S3 iD ID S3 CD Si ID S3 Si S3 iD S3 iTl ID iD iD S3 CD S3 Si S3 CD ■D lA.1 iD S3 S3 CD S3 SJ S3 IP IT I cn i c 1 I S3 I ■rt !■>- GD T$- ■x IJLI i ~ iT 1 *-H CO 'J3 m vr ro \r> ['-; ! S3 in r*3 ■— > ITl IV in IT' i CO 00 CD CO rv IV i v - r i IX I APPENDIX II i A PROBABILISTIC FORMULATION OF THE NON-UNIQUE ACCESS PROBLEM In this appendix, we describe the theoretical basis of the probability model discussed in Section 5. The problem is to calculate the prob- ability that a given retrieved record which has been accessed by a combination of retrieval keys such as name, date of birth, etc., will indeed correspond to an individual whose record is requested. It is to be noted that this theoretical development is completely independent of the retrieval algorithm itself. It supposes that a certain record has been retrieved and that it has certain characteristics which match characteristics known by the inquirer. The question is then: what is the probability that it is in fact the record desired by the inquirer? Note that there is some conceptual difficulty in even defining precisely when a record is, in fact, the record of a particular person. This is because the record may not uniquely characterize a person or may contain erroneous information. There are also attributive values which carry more weight than others in identifying a record-person correspondence (e.g., name). For purposes of the discussion, however, we assume that it is meaningful to say that a record matches or corresponds to an individual, i.e., js_ the record of the individual. This proposition or event we designate as 'M. '; that is, M. = a given record matches the individual i /,\ (e.g., who has attributes name = x,, address = x«, ..., attribute. = x.) The second event to be specified in the probabilistic formulation is that a record contains a particular attribute value for the jth retrieval key: K. = a given record has a particular attribute (2) va lue for key j (e.g., name = x,) It is useful to define a third "reference" event that defines the universe of discourse or context of the probability formulation. We call this A. In what follows we can consider A to be the selection of a given record from a data bank where the mode of selection is not specified. The basic probability to be calculated is then P(M. |Ki & K 2 & ... & K n & A) (3) 174 where '&' indicates logical conjunction. Thus (3) is the probability that a given record matches the individual i conditional on the fact that the record contains certain attribute values on the set of keys j = 1, ..., n. The basic problem of non-unique access is to select a set of keys so that the probability (3) is near unity. Let us first solve the problem of calculating probability (3) by study- ing the case of a single key, i.e., the probability P(M 1 |K J & A) (4) By the multiplication principle we have P(M, |A) P(K, |M, & A) P(M.|K, & A) = 3 J-J (5) i ' J P(K,|A) For readability we will omit the reference event A, thus (5) assumes the simpler form. P(M.) P(K-|M.) P(M, |K.) = 3 LJ_ (6) i ' J P(K.) The factors on the right have the following meaning: P(M-) = the a priori probability that a (7) 1 randomly selected record matches the individual i P(M. ) = 1/N if N is the total population upon which the data bank is drawn.* For example, if the individual i is known to be a California driver and the data bank contains all and only California driver records, then N = the size of the data bank. On the other hand, if the data bank concerns information on adult U.S. citizens, then N = the number of adult U.S. citizens. * In the complete formulation with the reference event A, A could be con- sidered to contain the information regarding the population upon which the data bank is based. 175 We next have: P( K - | M. ) = the probability that a record contains (8a) a specific value for retrieval key j conditional on the fact that the record matches the individual i. For example, if the j th key is i's address, then P(K. |M. ) is the prob- ability that i's record contains his correct address. It therefore follows that P( K • | M . ) = the complement of the sum of the error (8b) rate in key j and the relative frequency of omission of data in key j Note that for perfectly accurate records, we would have P(K- |M, ) = 1.* The third factor P(K. ) is estimated as the relative frequency of the specific information given by key j in the bank. For example, if two records are retrieved having the same value for key j and m is the size of the data bank then P(K-) is estimated as 2/m. Now if a set of keys j = 1, ..., n are independent , the probability (3) is calculated by putting the conjunction K & ... &K n for K- in (5) (or the abbreviated Form (6)) and expanding the result into a product for j = 1 , . . . , n ; i.e., n P(K.|M.) P(M. K,& ... &KJ = P(M.) IT J J (9) However, the independence assumption is only valid for certain keys. For example, name and birthdate can be regarded as independent, but not birth- date and year of graduation from high school. Independence also depends on the conditional probability. We therefore must look for a more general solution. k The error rate in query formulation is not part of the probability (8)> which has to do with the probability of the selection procedure being successful conditional on an error-free query. 176 One approach to this problem is to use some probability machinery developed by W. E. Johnson and reported in Keynes.53 There are also versions of the Johnson work in Carnap. 5 ^ We now present our version of the so-called "Johnson Cumulative Formula." A set of relevance coefficients or coefficients of influence is first defined. For any propositions (events) K, , . . . , K , we define P(K,|K ? & A) C(K,,KjA) = i —^ (10) 1 2 P(K X |A) C(K r K 2 ..., KJA) = C(K 1$ K 2 |A)-C(K 1 & K 2 , K3, ..., KjA) (11) It is then easy to show that C(K, , ..., K |A) has the operator property* P(Kj & ... & K n |A) = C(K r ..., KJA) JT P(Kj|A) (12) (see Carnap55) # >i J The coefficient (11) is thus a coefficient of independence and is unity if the events K, , K«, . ..» K are independent relative to the event A. The cumulative formula is obtained as follows: Let K be the conjunction K,& ... &K . Then by the multiplication principle P(M. |A) P(K|M, & A) P(M.|K&A)= ] (13) 1 P(K|A) Applying (12) to the numerator and denominator of (13) and then multiply- ing both sides of the result by [P(M i fA) ] n "l we get * We use our own notation for the coefficient. 53 Keynes, J. M. , A Treatise on Probability , MacMillan and Co., London, 1929, pp. 149-155. 54 Carnap, R. , The Logical Foundations of Probability , University of Chicago Press, 1950, pp. 356-360. 55 Ibid , p. 358, Formula T66-3C. 177 (14) n1 C(K,, ..., K n |M. & A) n / P(M. | A) P(K. |M & A) [P^A)]"' 1 • P(M.|K & A) = — i H_J JT a C(K r ..., KJA) j=l \ P(Kj |A) We now use the cumulative formula (14) in the following way. First intro- duce the abbreviations. V = P(M.|K & A) (15) P = P(M.|A) (16) q j = P(M i |K j & A) (17) By the multiplication principle, q. is the factor in the parentheses of (14), i.e., J P(MjA) P(K.|M. & A) q, = ] J — ] (18) J P(Kj|A) Our problem is to study the expression for V. Using the abbreviations, the formula (14) becomes . C(K,, ..., K |M. & A) n P n_1 v = — i Q— ! TT q, (19) C(K X , ..., K n |A) i-1 J We now use (19) again, but this time substitute '^M. ', i.e., the negation of M-j , for 'Mj' in (14). The quantities p, V, and qj are then replaced by their complements. We get. n 1 C(K , ..., K KM. & A) n (1-P) n_1 (1-V) - — i- n —± i TT (1 - q.) (20) C(K p ..., K A) j=l J 178 The ratio of (20) to (19) yields --CfsR 1 ) where E is the factor C(K,, ..., KjoM, & A) E = i 1 (22) C(K r ..., K | M - & A) The equation (21) is the solution to our problem; that is, the right side of (21) is the product of the two quantities, E as defined by (22), and so that (21) becomes from which we derive 7 - I'" EF v_I 1 ♦ EF < 24 > We see that V depends not only upon the quantities q. but also on p, the a priori probability of M. , and the quantity E. The J value used for E is usually obtained by independence assumptions . Note that if E < 1 then, from (24), 179 V 1 + F (25) This means that a lower limit to the probability of M- conditional on the combination K, & K~ & ... & K can be computed if E < 1. The value E = 1 occurs when the numerator and denominator of (22) are equal. In particular, this is the case when the keys are independent relative to both~M- and M. . The smaller the value of E, the greater the ability of the keys to 1 discriminate. To summarize: We have three formulas for the desired probability V each involving a different ratio of coefficients of influence: the formula (19), the formula (20), and the formula (24). In Section 5 we have implemented the formula (24) for a range of values of E. 180 APPENDIX III SOFT-MATCH TECHNIQUES In Section 5.2, we noted the classical problem of trade-off between re- call and precision, i.e., increasing recall lowers precision. But this arises only in the case where all retrieved records are treated on an equal par and not subjected to further filtering. Our approach to the problem is to increase the recall and then to provide a mechanism for additional filtering to improve the precision. In this section we dis- cuss a method for increasing the recall called "Soft-Matching." The result of a Soft-Match Technique is not only a set of retrieved records, but also a numerical weight assigned to each record to indicate its closeness to the query. As an example, consider the case of a query which is a conjunction of three key values denoted by A, B, C. A tradi- tional retrieval would select only records that matched on all three. A soft-match technique might be to also select those records that matched on A and B but not on C, on A and C but not on B, and on B and C but not on A. These secondary retrievals would then be assigned weights less than that assigned to a record that matched on all three. The retrieval could also be extended to tertiary retrievals such as records that match- ed on A but not on B and not on C, etc. These would be given the lowest weights. In this appendix a heuristic weighing scheme is presented for filtering the candidate records. In Section 5.5 and Appendix II we presented a second weighting scheme based on probabilities which is in fact the measure of precision for a single retrieval. (That is to say, precision is the proportion of retrieved documents that are relevent, i.e., match the inquirer. Consequently, if exactly one record is retrieved, then the precision corresponds exactly to the probability that the record is in fact that desired by the inquirer.) On the basis of the above discussion, we see that a soft-match technique involves two components: (1) an algorithm for determining a subset of the data base of manageable size which has maximum recall; (2) a scoring function for assigning the closeness measures to each item in the subset. First, consider the problem of determining a manageable subset. Suppose that the retrieval prescription involves n keys K| , K« j ...» K Then let S^) be the set of records which match on the i th key. In principle, the subset with maximum recall would be the set union of S(K,), •••» S(K n ). This approach would not be efficient however, so a method which utilizes the error and omission rates and the discrimination factors is recommended. These quantities are defined as follows: 181 J.L. a. = probability that the key value for the i key is incorrect b. = probability that the key value for the i key is omitted e i ■ a i + b i (Thus, the complement of e. is the quantity c. defined in Section 5.1). The discrimination factor of a key is measured by the relative frequency of a key value in the data base. Let R. be the relative frequency for the i th key. This quantity R. is estimated for certain keys by the values given in Tables 5.3-1, 5.3-2; it therefore corresponds to the concept of retrieval ratio. It follows that the initial selection of the retrieval should be the union of, not all n sets S(K.), but a certain selection of fewer sets say S(K. ), ..., S(K. ) (m (i.e., a partially matching key) should: 182 (1) Increase as R. decreases (2) Increase as f. increases (3) Increase as a. (the error rate in the i key) decreases. It is therefore plausible to take the contribution of the i key to the total score (in the case that f . > 0) to be proportional to the quantity f i R. l since this is a simple expression that satisfies the requirements (1), (2), and (3). It is convenient to normalize the scores, however, so we will take the constant of proportionality to be 1/R where n 1 R = z -4- (l) i=l i Thus, the score for the i key is 1 f i (1 " a i } S = — — — (2) *1 R R. {d) We now define the total score as consisting of two components: a positive component made up of the contributions S. defined above and a negative component to be defined below. ] The positive component we take to be simply ; i (3) n S = I i=l (Note that this summation runs over all the keys in the query, and not over just the subset of m keys used to get maximum recall.) By (1) and (2) we see that S is a convex linear combination of the terms f.(l-a.) where i = 1, ..., n, and so S can range from to 1. To define the negative component, we see that if a key is error free and fails to match (f. = 0), then it should kill off the entire retrieval. Thus, in the case f . = we should have a negative contribution which 183 varies as (1 - a.)- We therefore take the score when f. = to be i l V-(l-a.) Since a value of T.= -1 for only one key is sufficient to invalidate a retrieval, we take the total negative score to be, not the sum of the individual scores, but rather the negative of the maximum value of (1 - a.), i.e. , T = - max (1 - a. ) i f.=0 Thus T ranges from -1 to 0. The retrieval scoring function is thus U = S + T An example of how this scoring function is used is given in Table III-l The example deals with the scoring of four records based on values of the parameters R. (the retrieval ratio) and a. (the error rate). 184 Table III-l. Example of Soft-Match KEY NAME ADDRESS SEX R i .003 .00002 .5 a i .001 .01 Table of Values for R. , a. NAME ADDRESS SEX Record 1 match (1) match (1) no match (o) Record 2 match (1) partial match ( .75)* match (1) Record 3 partial match (.83)** match (1) match (1) Record 4 match (1) match (1) data absent Table of Retrieval Outcomes (values in parentheses are those for f. ) S T U Record 1 .9900 -1 -.0100 Record 2 .7442 .7442 Record 3 .9889 .9889 Record 4 .9900 .9900 Retrieval Scores * The partial match example is for the case that one character out of four doesn't match. ** The partial match example is for the case that one character out of six doesn't match, 185 APPENDIX IV FORTRAN SOURCE CODE FOR THE PROBABILITY MODEL 186 FORTRAN IV D06-31 SOURCE LISTING PAGE 001 8001 0082 8883 8004 0885 8886 8887 8088 0009 00)0 08] i 0012 80 11 C QUERY SIMULATES THE DATA BASE/USER INTERACTION DIMENSION A< 1 1) .B(ll) ,CC 1 1),RC i 1),S< 11) DIMENSION JARYCi 1), IFLD(ll) DIMENSION ARGS CI 1) DIMENSION SURNMC 18) COMMON IRAN. JRAN, EARYC27) , PARYC3) DATA ARG3 ( 1 ) /. 23972SE-2/, ARGS C2) /. 8333 DATA ARGS C 4) / . 02/, ARGS /. 0002/ 00Q333333/. ARGS ( 18) /. 555555E-4/ SURNMC4)/.98?E-?, . 2459E-5, .9635E- 2 , .272037E-4/ DATA SURNMC 5) DATA SURNMC 8) DATA DATA ,SURNMC6).SURNMC7)/-4.81966E-5,S,627E-5, 1.265E-4' . SURNMC 9) , SURNMC 1 B) /4.8 197E-4, 7 . 832E-4, . 3384E-2- P AR Y CI), PAPY (2) , PARY C 3 ) /4 . 545454E-9, 2 . 5E-7 , 1 . E -6- E AR Y k 1 ) , EAR Y< 2 ) , EAR Y C 3 ) , E AR Y C 4) / . 5 , 1 . , 2 . , 1 , / Q >K>K>t<>t"+ ■■■*■*■ H:***********^ c QUERY SIMULATES QUERY PROCESSING AGAIN; IT GENERATES VALUES FOR ACJ)CTHE ERROR . I A DA l A BASE RATE IN A FIEt ,D3 BCJ) CTHE OMISSION PATE OF DATA IN A FIELD] C C J ) L 1 - C A C J ) +0 C J ) ] . . . THE COHF I DENCE FACTOR RCJ) THE PERCENTAGE OF THE FILE RETRIEVED VIA A FIELD, AND SCO A QUERY SCORE, WHICH RANGES FROM 1 TO THE FILE SIZE. SCJ) IS NOMINALLY THE RATIO CCJ)/RCJ), AND IS ADJUSTED HERE TO MAINTAIN REASONABLE LIMITS. THE USER ENTERS THE NUMBER OF ARGUMENTS CFROM 1 UHICH HE LUSHES TO USE IN HIS QUERY. THE QUERY IT REQUIRED A SUCCESSFUL BOOLEAN AND OF ALL THE SUCCESSFUL. THE PROGRAM LULL TERMINATE WHEN THE NUMBER OF INPU IS SET TO ZERO. TO 11) IS TREATED AS ARGUMENTS TO ARGUMENTS IF BE FOR THIS TEST THE VALUES OF A C.J) AND BCJ) WILL BE TREATED AS ZERO, FOLLOUING THE ENTRY OF THE FIRST NUMBER— THE NUMBER OF ARGUMENTS TO FOLLOU--THE USER MAY ENTER NUMBERS CORRESPONDING TO THE ACTUAL ARGUMENTS. EACH ARGUMENT HAS AN ASSOCIATED RCJ), UHICH IS THE PERCENTAGE RETRIEVAL TO BE EXPECTED. THE ARGUMENT NUMBERS AND THEIR ASSOCIATED VALUES ARE AS FOLLOWS: C ARGUMENT NUMBER MEANING RETRIEVAL RATIO 0814 C C 9 C 18 C BEGIN BY ASKING FOR 1 JRITE (6,285) C NOW EXAMINE HARGS TO DATE OF BIRTH CMMDD) MONTH OF BIRTH YEAR OF BIRTH STATE SEX STREET NUMBER STREET NAME EDUCATION COUNTY DATE OF 1/365 1 / 1 2 1/50 1/58 1 / i , i /5 , e 1 / 1 5 i/3800 BIRTH CYMD) OPERATOR INPUT. START THINGS GOING. 000 00 C EST I MATED) 1/18,098 187 FORTRAN IV D0G-01 SOURCE LISTING PAGE 002 £ X; >f~K ?t "S' H< >H t< ?f- >K >K ?fc sfof« 5fc ■+-' ^foto+of- ' 00 i 5 READ (6,216) HARGS 00 16 If- '' NARGS) 160, 100, 10 C COME HERE TO PREPARE A QUERY SURROGATE. 06 J ? 1 DO 20 I = i . NARGS 0016 WRITE (6,220') 0019 READ (6.210) J 0020 A ( I) =0 . 0021 B( I) =8 . 0622 C(I)=1. 0021 R ( I ) =ARG5 (J) 0024 9( I) =f"C I) /R( I) ) 0025 JARY(I)=J 002b 20 CONTINUE 0027 WRITE (6,2i5) C NOU CALL ON THE MAIN COMPUTATIONAL ROUTINE TO PREPARE THE TABLES JARY(HARGS+i) =1 1 H=NARCP+1 UP 1 TE ( 5 , 2O0 ) NARGS DO 25 J=l, NARGS WRITE (5, 245) J, A (J) , J, B ( J) , J,C( J) , J,R ( J) , J,SC\J) , J, JARY(J) CONTINUE DO 50 1 = 1 , 10 COHARGS+l 1 ' = 1. B (NARGS+1 ) =0. A(NARGS+1 ) ~3 . R(HARGS-M) =SURNM( I) SCNARGS+1) =C(HhRGSM)/R(NARGS+1) WR ITE(5,250) DO 45 J=i .NARGS GOTO < 3 1 .■ 32 .- 33 , 34, 35 , 36 , 3? - 30 , 39 , 40 > JAPY( J ) WRITE (5, 255) GOTO 45 URITE(5, 260) GOTO 45 WRITE (5, 265') GOTO 45 WRITE (5, 276) GOTO -45 WRITE (5, 275) GO I 45 UR I TEC 5 .2SQ) GOTO 45 LIR ITE(5,285) GOTO 45 WRITE (5, 290) GOTO 45 WR ITFC5.. 295) GOTO 45 WRITE (5, 360) CONTINUE WRITE (5, 305) WR ITE(5, 150) WRITE (5,225) I WRITE (5, 230) 188 0028 0029 0030 007 1 0032 0033 25 0O34 0035 0036 0037 0038 0039 0040 0041 0042 0043 31 0044 0045 32 0046 O047 ~k"Z 48 O042 34 0050 005 1 35 0052 0053 36 0054 0055 ~\~ : ' 6056 0057 *iH O058 0059 39 0060 Q06 1 40 OOfc? 45 O063 0064 0G65 OO G6 FORTRAN IV DOb-Sl SOURCE LISTING PAGE 003 0667 6068 0869 0070 037 1 6072 0073 0074 0875 0876 0077 0078 0073 0080 WRITE' 5.. 235) WRITE (5,248) CALL P VC Y( ft , S . C , R , S.. N , JAR Y) 58 CONTINUE GO TO i 100 CALL PVCY(A,B*C,R,S,HARGS, JARY) CALL EXIT C FORMAT STATErENTS 153 FORMAT (1H , i3X, 1 10 ('-'). 7X) 200 FORMAT ( ihi . 'NARGS=' , 12) 205 FORMAT ('8 ENTER NUMBER 213 FORMAT (12) 215 FORMAT (' EXITING QUERY") 220 FORMAT (* ENTER QUERY FIELD IDENTIFIER') 225 FORMAT ( 1H , 13X, * ! ' , 108X, ' ! ' , 7X/1H , i3X, * 1 = SURNAME DECILE' , iX, 12, 4SX, ' ! * , 7X) OF ARGUMENTS '•'} 43 X, 0081 230 FORMAT'. 1H , 13X, ' !' , iOBC-' ) , ' ! " ,7X/ 1 iH » i3X, ' J * , 3X, ' ERROR RATE ' , 2X, ' !' , 2X, * RECALL FACTOR' , 2 2X, ' ! •' ,28 X, 'FILE SIZE',37X,' !' ,7X) O082 235 FORMAT ( iH . i3X, ' ! ' , ibX, ' ! ' , i7X, ' !* ,74('-' >,* ! * , 0084 245 FORMAT( iHO . 'AC , 12, ' ) =' , tib.8,3X, 'BC , 12, ' ) =' ,hib.3. l 3X, ' C (' , I 2, ■ ) =' ,E15.8,3X, ■RC , 12, ' ) =' ,E15.8,3X, ' S c. ' , 2 12, ' ) ** .E15.8..3X, ' JARYC , 12, ' ) =' , 12) 0035 250 FORMAT ( 1H1 1 FIELDS') ,43X, 'PRECISION OF RETRIEVED RESULTS FOR DATA 0086 255 FQRMATC IH , 48X, 'DATE OF BIRTH (MONTH AND DAY ONLY)') 0087 260 FORMAT ( iH , 48 X , ' MO hi TH OF B I R TH ' ) 0088 265 FORMAT ( IH , 48X, ' YEAR OF B IRTH * ) 0089 278 FORMA T(1H ,48X, 'STATE') 8896 275 FORMAT( IH ,48X, 'SEX' ) 009 1 260 FORMAT ( lH ,48X, 'STREET NUMBER') 6692 295 FOPMAT( IH ,48X, "STREET NAME' ) 0093 290 FORMAT ( IH ,4BX, 'EDUCATION' ) 0094 295 FORMAT ( iH , 48X, ' COUNTY' ) 0095 300 FORMAT''. iH , 48X, 'DATE OF BIRTH (YEAR, MONTH, AND DAY)''' 0096 365 FORMATC iH ,48X, 'AND SURNAME'' 0097 END 189 FORTRAN IV STORAGE MAP A 008886 REAL #4 ARRAY (It) B 088062 REAL*4 ARRAY (11) C 6013136 REAL*4 ARRAY (11) R 888212 REAL #4 A R R A Y (11) S 888256 REAL-K ARRAY (11) J A R Y 8 8 3 42 I N TEG E R *2 A R R A Y (11) IFLD 888378 INTEGERS ARRAY (11) ARGS 888416 REAL+4 ARRAY (11) SURNM 008472 REAL>k4 ARRAY (18) H A R G S 8 Q 2 1 8 I H TE G E R *2 VA R 1 AB L E I 882 i 82 INTEGERS VARIABLE J 002184 INTEGERS VARIABLE H 882106 IHTEGER*2 VARIABLE PVCY 008088 REAL*4 PROCEDURE EXIT 080888 REAL +4 PROCEDURE COMMON BLOCK // LENGTH 000174 INTEGERS VARIABLE IHTEGER*2 VARIABLE REAL*4 ARRAY (27) REAL*4 ARRAY (3) IRAN 680008 JRAN 080882 EARV 888804 PAR'r' 888 160 190 FORTRAN IV DB6-81 SOURCE LISTING PAGE 081 C MAIN COMPUTATION ROUTINE OF PRIVACY .MODEL—MARCH 1976 **NOTE: THIS SUBROUTINE HAS BEEN MODIFIED TO PRODUCE A TABULAR PRINTOUT. 803 i SUBROUTINE PVCY(AJ . BJ, CJ, RJ, SJ, NARGS.. JARY) C DECLARE VARIABLES 62 D I ME N S 1 hi J A R Y ( 1 D 0903 DOUBLE PRECISION B IGPH 1 15 .F,VARYC 13. 3) 0004 COMMON J RAN , -IRAN .. EAR Y( 27 ) , PARY ( 3 ) 0005 REHL AJ ( 1 1 ) , BJ ( 1 1 ) , CJ Cl 1) , RJ ( 11 ) , SJ ( 1 1 ) - P 0006 REP 1 . QARYcii) O 7 D ATA VA R Y i 1 , 1 ) . '9 . D 6 / |J^h>KJK?K?fC>',<.'-;'-~tc:i<>fiH;?f-^>l'itK;»< C MAIN LOOP STARTS HERE C C VALUES ARE RETURNED IN THE ARGUMENT ARRAYS— TERMINATE IF NARGS< = 8 OOOS IF(NAPGS) 95,95.. 11 C BEG 7 N PROCESSING THE QUERY RESPONSE HERE l ^U09 i I CONTINUE H:j 10 DO 90 JK" i .. 13 00 11 L=.TK 00 1 2 CONF =1. _ 0. 085* C L - 1 ) 00 1 3 PEF AC =C ONE** ( NARGS ) 80 i4 ERRAT= I . -CONF 00 i 5 PERERR=ERRAT*l88 0015 DO SO I - i , 3 0*17 P=PHKY(I) 08 i 6 DO 20 1 1 = i , NARGS 88 1 9 20 OAR Y ( I J ) =p*S J ( I I ) *€ONF C NOU COMPUTE THE FUNCTION F C>^*?M ! :>t<>M!>i<>lo|«ts>M©(<>|0|©t<>t<>fc 0020 F-l. 002 1 X=P/< i . -F) 8822 DO 30 JJ=1 -NARGS 9823 BIGPI f JJ"' = ( 1 .-QARYCJJ; >/DARY(JJ) 00 24 3 G F = F *>v KB I G P I (J J > 8825 F=F/X C PREPARE THE PRINTOUT TABLE B 2 S V A R Y ( J !< .• I ''' = i . " ( 1 . +F ) 3027 IF(VARYCJK, T) .GT. (l.On VARYCJK, I) = 1 . 8 dy^y 80 i.:ON i INUE 0030 URITE<5, 150) PERERR, REFAC. ?-:-'i¥W ¥■''«>¥. 0034 9 5 UR IT E C 5 , 288 ) 8835 RETURN 191 FORTRAN IV D86-01 SOURCE LISTING PAGE 802 C F 'MAT 6 TATE MEN T SECTION 8036 t-riS FORMAT UH , i3X, ' i ' , 15X, ' f ' , 17X, * I " , 23X. ' .{ ' , 27X, ' ! * , 22X, ' ! ' , ?XS 1. !H , )3X, ' ! ' ,4X,FG.2, **' , 4X., ' ! ' ,4X,F8.4,5X, ' ! '* , 4X, ? Ei5.8,4X, " i ' ,8X,Eib.8,6X, ' i ' , 3X, E 15 . 8,4X, ' ! ' , 7X) 0637 i70 FORMATUH .- i3X, i 18 ( ' -* ) , 7X) £1038 260 FORMAT <1 HO,' NORMAL TERMINATION') 0038 END 192 FORTRAN IV STORAGE MRP .JhPV 668838 INTEGER*2 PARAMETER ARRAY (11) BIGPI y&0832 VARY 060 i 62 AT 8886 i4 B.J 6868 ib CJ 888020 RJ 888022 3 J 1*60824 QARY ■366652 NARGS 8 nm? 2 6 F 06 L 140 P 801 150 JK 601 154 L 081 i5b COHF 001 166 REFAC 86 i 164 ERRAT 06 1 176 PERERR 661 174 I 66 i 266 II 66 i^e^ X 06 i 264 J J 00 1210 REAL*8 ARRAY (.11) i REAL*8 ARRAY (13, , 3) VECTORED REAL+4 PARAMETER ARRAY an REAL*4 PARAMETER ARRAY (11) REAL*4 PARAMETER ARRAY (11) REAL*4 PARAMETER ARRAY i i i ) REAL*4 PARAMETER ARRAY (11) REAL ^4 ARFA ■:" ( i i) INTEGER*2 PARAMETER VARIABLE REALMS VARIABLE REAL ^4 VARIABLE INTEGER*2 VARIABLE INTEGERS VARIABLE REAL*4 VARIABLE REAL*4 VARIABLE REAL*4 VAR I ABLE / REAL*4 VARIABLE INTEGER*2 VARIABLE IHTEGER'i'2 VARIABLE REAL*4 VARIABLE INTEGER*2 VARIABLE :OMMON BLOCK // LENGTH 086174 INTEGER*2 VARIABLE I NTEGER*2 VAR I ABLE REAL*4 ARRAY (27) REALK4 ARRAY (3) ERAH 089890 JRAN 8W8882 EARY 666004 PARY 606166 193 NBS-114A (REV. 7-73) U.S. DEPT. OF COMM. BIBLIOGRAPHIC DATA SHEET 1. PUBLICATION OR REPORT NO. NBS SP-500-2 2. Gov't Accession No. 3. Recipient's Accession No. 4. TITLE AND SUBTITLE Accessing Individual Records From Personal Data Files Using Non-Unique Identifiers 5. Publication Date February 1977 6. Performing Organization Code 7. author(S) Gwendolyn B. Moore, John L. Kuhns, Jeffrey L. Trefftzs. Christine A. MontnorrexY- G ORGANIZATION NAME ANlD ADDRESS 8. Performing Organ. Report No. 9. PERFORMING Operating Systems, Inc. 21031 Ventura Boulevard Woodland Hills, California 91364 10. Project/Task/Work Unit No. 640.1118 11. Contract/Grant No. 5-35928 12. Sponsoring Organization Name and Complete Address (Street, City, State, ZIP) National Bureau of Standards Department of Commerce Washington, D.C. 20234 13. Type of Report & Period Covered Final 31 March 1976 14. Sponsoring Agency Code IS. SUPPLEMENTARY NOTES Library of Congress Catalog Card Number: 76-57950 16. ABSTRACT (A 200-word or less (actual summary of most significant information. If document includes a significant bibliography or literature survey, mention it here.) The Privacy Act of 1974 places restrictions on the Federal, state and local agencies 1 use of the Social Security account number as an identifier. For some agencies, compliance will involve changes in implementation of retrieval algorithms. This report describes methodology applicable to these changes in the more general context of the problem of retrieving individual records from files using non-unique identifiers. State-of-the-art retrieval techniques are discussed, a method for assigning reliability weights to various personal data elements is presented, file validation techniques for the error and omission rates of data items are suggested, and a retrieval probability model — designed to show likelihood of retrieval of a subject's record given a variety of populations, combinations of identifiers, and error/omission rates--is described. A methodology is developed for forming confi- dence factors from the established error/omi'ssion rates for combinations of non- unique identifiers that are candidates for use as retrieval keys. Use of these confidence factors as indices into the precision tables produced by the probability model is described. 17. KEY WORDS (six to twelve entries; alphabetical order; capitalize only the first letter of the first key word unless a proper name; separated by semicolons) Data retrieval; file validation; name lookup; non-unique identifiers; personal data files; Privacy Act; probability model; retrieval. 18. AVAILABILITY f)(~l Unlimited ! For Official Distribution. Do Not Release to NTIS l_2f Order From Sup. of Doc, U.S. Government PjLntuift^ftffi uraer r rom sup. ot uoc, u.;>. Government fjLQtu3a,utttce Washington, D.C. 20402, SD Cat. No. C13 « i-V'.QVV-d _J Order From National Technical Information Service (NTIS) Springfield, Virginia 22151 19. SECURITY CLASS (THIS REPORT) UNCLASSIFIED 20. SECURITY CLASS (THIS PAGE) UNCLASSIFIED 21. NO. OF PAGES 2 03 22. Price $2.65 USCOMM.DC 29042-P74 »U.S. GOVERNMENT PRINTING OFFICE: 1977 240-848/120 1-3 ANNOUNCEMENT OF NEW PUBLICATIONS ON COMPUTER SCIENCE & TECHNOLOGY Superintendent of Documents, Government Printing Office, Washington, D. C. 20402 Dear Sir: Please add my name to the announcement list of new publications to be issued in the series: National Bureau of Standards Special Publication 500-. Name Company. Address _ City State Zip Code (Notification key N-503) NATIONAL BUREAU OF STANDARDS The National Bureau of Standards 1 was established by an act of Congress March 3, 1901. The Bureau's overall goal is to strengthen and advance the Nation's science and technology and facilitate their effective application for public benefit. To this end, the Bureau conducts research and provides: (1) a basis for the Nation's physical measurement system, (2) scientific and technological services for industry and government, (3) a technical basis for equity in trade, and (4) technical services to promote public safety. The Bureau consists of the Institute for Basic Standards, the Institute for Materials Research, the Institute for Applied Technology, the Institute for Computer Sciences and Technology, and the Office for Information Programs. THE INSTITUTE FOR BASIC STANDARDS provides the central basis within the United States of a complete and consistent system of physical measurement; coordinates that system with measurement systems of other nations; and furnishes essential services leading to accurate and uniform physical measurements throughout the Nation's scientific community, industry, and commerce. The Institute consists of the Office of Measurement Services, the Office of Radiation Measurement and the following Center and divisions: Applied Mathematics — Electricity — Mechanics — Heat — Optical Physics — Center for Radiation Research: Nuclear Sciences; Applied Radiation — Laboratory Astrophysics " — Cryogenics 3 — Electromagnetics s — Time and Frequency 2 . THE INSTITUTE FOR MATERIALS RESEARCH conducts materials research leading to improved methods of measurement, standards, and data on the properties of well-characterized materials needed by industry, commerce, educational institutions, and Government; provides advisory and research services to other Government agencies; and develops, produces, and distributes standard reference materials. The Institute consists of the Office of Standard Reference Materials, the Office of Air and Water Measurement, and the following divisions: Analytical Chemistry — Polymers — Metallurgy — Inorganic Materials — Reactor Radiation — Physical Chemistry. THE INSTITUTE FOR APPLIED TECHNOLOGY provides technical services to promote the use of available technology and to facilitate technological innovation in industry and Government; cooperates with public and private organizations leading to the development of technological standards (including mandatory safety standards), codes and methods of test; and provides technical advice and services to Government agencies upon request. The Insti- tute consists of the following divisions and Centers: Standards Application and Analysis — Electronic Technology — Center for Consumer Product Technology: Product Systems Analysis; Product Engineering — Center for Building Technology: Structures, Materials, and Life Safety; Building Environment; Technical Evalua- tion and Application — Center for Fire Research: Fire Science; Fire Safety Engineering. THE INSTITUTE FOR COMPUTER SCIENCES AND TECHNOLOGY conducts research and provides technical services designed to aid Government agencies in improving cost effec- tiveness in the conduct of their programs through the selection, acquisition, and effective utilization of automatic data processing equipment; and serves as the principal focus within the executive branch for the development of Federal standards for automatic data processing equipment, techniques, and computer languages. The Institute consists of the following divisions: Computer Services — Systems and Software — Computer Systems Engineering — Informa- tion Technology. THE OFFICE FOR INFORMATION PROGRAMS promotes optimum dissemination and accessibility of scientific information generated within NBS and other agencies of the Federal Government; promotes the development of the National Standard Reference Data System and a system of information analysis centers dealing with the broader aspects of the National Measurement System; provides appropriate services to ensure that the NBS staff has optimum accessibility to the scientific information of the world. The Office consists of the following organizational units: Office of Standard Reference Data — Office of Information Activities — Office of Technical Publications — Library — Office of International Relations — Office of International Standards. 1 Headquarters and Laboratories at Gaithersburg. Maryland, unless otherwise noted; mailing address Washington, DC. 20234. - Located at Boulder, Colorado 80302. U.S. DEPARTMENT OF COMMERCE National Bureau of Standards Washington. DC. 20234 OFFICIAL BUSINESS Penalty for Private Use. $300 PENN STATE UNIVERSITY LIBRARIES A0DDD711 2EM5M U.S. DEPARTMENT OF COMMERCE COM-215 u w\ LS.MAIL SPECIAL FOURTH-CLASS RATE BOOK