A Celebration of Statistics The lSI Centenary Volume International Statistical Institute ~ Editorial Advisory Panel B. A. Bailar, Bureau of the Census, Suitland, MD, U.S.A. J. M. Chambers, Bell Laboratories, Murray Hill, NJ, U.S.A. A. P. Dawid, University College, London, England I. Guttman, University of Toronto, Toronto, Ontario, Canada R. A. Horvath, Central Statistical Office, Budapest, Hungary D. Vere Jones, Victoria University, Wellington, New Zealand A Celebration of Statistics The lSI Centenary Volume A Volume to Celebrate the Founding of the International Statistical Institute in 1885 Edited by Anthony C. Atkinson and Stephen E. Fienberg Springer-Verlag New York Berlin Heidelberg Tokyo ANTHONY C. ATKINSON Imperial College of Science and Technology London SW7 2BZ, England STEPHEN E. FIENBERG Carnegie-Mellon University Pittsburgh, Pennsylvania 15213 U.S.A. Library of Congress Cataloging in Publication Data Main entry under title: A Celebration of statistics. Includes index. I. Mathematical statistics-Addresses, essays, lectures. 2. Statistics-Addresses, essays, lectures. I. Atkinson, A. C. (Anthony Curtis) II. Fienberg, Stephen E. Ill. International Statistical Institute. QA276.16.C451985 519.5 84-26825 © 1985 by Springer-Verlag New York Inc. Softcover reprint of the hardcover 1st edition 1985 All rights reserved. No part of this book may be translated or reproduced in any form without written permission from Springer-Verlag, 175 Fifth Avenue, New York, New York 10010, U.S.A. The use of general descriptive names, trade names, trademarks, etc., in this publication, even if the former are not especially identified, is not to be taken as a sign that such names, as understood by the Trade Marks and Merchandise Marks Act, may accordingly be used freely by anyone. Typeset by Aseo Trade Typesetting Ltd., Hong Kong. 987654321 ISBN -13: 978-1-4613-8562-2 e-ISBN-13:978-1-4613-8560-8 DOl: 10.1007/978-1-4613-8560-8 Foreword The International Statistical Institute was founded in 1885 and is therefore one of the world's oldest international scientific societies. The field of statistics is still expanding rapidly and possesses a rich variety of applications in many areas of human activity such as science, government, business, industry, and everyday affairs. In consequence, the celebration of the Institute's centenary in 1985 is of considerable interest not only to statisticians but also more widely to the international scientific community. As part of its centennial celebration planning the Institute decided to publish a volume of papers representing the immensely wide range of interests encompassed by statistics in its international context, viewed both from a historical and from a contemporary standpoint. We were fortunate in securing the services of Anthony Atkinson and Stephen Fienberg as Editors of this volume: they have worked hard over a period of several years to put together a most fascinating collection of papers. On behalf of the Institute it is my pleasant duty to thank them and the authors for their contributions. J. DURBIN, President International Statistical Institute Preface The papers in this volume were prepared to help celebrate the centenary of the International Statistical Institute. During the lSI's first 100 years statistics has matured, both as a scientific discipline and as a profession, in ways that the lSI's founders could not possibly have imagined. With this maturation, not only has statistics become the language of science, but statistical terms and concepts have also permeated all aspects of modern society, from the workplace, to our news papers, and even to the courtroom. Being able to see statistics in action all around us is a source of constant stimulation to us as individual statisticians, and thus to our field. We hope that the papers in this volume reflect the variety, excitement, importance, and international range of our subject. There should be something here for everyone. When we wrote to prospective authors inviting papers we stated: "The plan of the International Statistical Institute is that the book should be truly international. We hope for a geographical spread of authors and also for a range of subjects that will interest statisticians in all countries, both developed and developing. We are most concerned that the papers should be a way of strengthening communication between groups of statisticians. In some cases an appreciable proportion of survey, exposition, and discussion may be more appropriate than new results. We are especially interested in a discussion of unsolved problems and of future challenges." At the back of our minds was the additional hope that many of the papers would provide a glimpse of what statisticians do--for laymen and for our colleagues in other professions. A brief outline of our method of working will help explain the contents of the volume. We received the invitation to edit the volume early in 1980. In the process of inviting and refereeing contributions we have had the support of a panel of advisors, listed on the first page, some of whom were chosen explicitly as representatives of the three sections of lSI. They and other members of the lSI were asked to suggest topics and names of authors for papers. As a result of these suggestions we wrote, in the autumn of 1980, to an initial list of 18 authors in 12 countries. Other invitations to contribute followed as the structure and needs of the volume emerged. Our purpose throughout has been to display the best of statistics, while at the same time achieving a balance across topics and across countries. We attempted to do this while relying upon authors whom we knew could provide interesting and even provocative papers. As the work on the volume progressed, the balance has shifted somewhat, in part by design and in part by happenstance. Some potential authors felt that they could not accept our invitation to contribute. Others accepted but then found that they could not produce the papers they had offered. In these cases, rather than lose the balance across subjects, we sometimes viii Preface had to turn to someone near at hand to produce a key paper. This process has slightly distorted the national balance for which we strove, but the quality of the papers has been our primary concern. Looking back over five years of work we are impressed by the quality, the quantity, and the variety of the papers which have been assembled. Some of the most technical papers are among the most readable and some of the less technical present us with the most fascinating challenges for the future. Nonetheless, we are aware of some major gaps which are worth mentioning if only in the hope that someone we did not contact will feel impelled to put pen to paper. One such gap is the difficult subject of the abuse, or perhaps distortion, of statistics by government and of the duties of government statisticians to their employers and to their fellow citizens. It is stated that Mussolini made the trains run on time. He is also said to have reduced unemployment in Italy by decreasing the categories of those counted as unemployed. Governments of many persuasions continue to behave in this manner, or at least occasionally attempt to do so. Some examples from British experience are quoted by Sir Claus Moser in Section 3, "Integrity," of his presidential address to the Royal Statistical Society, pub lished in 1980. Readers in other countries can doubtless provide their own examples. What we sought, but failed to find, was a paper that would give examples of this behavior in an international context. There are other ethical issues going beyond integrity to the statistician's responsibility for the use and interpretation of government statistics, especially those subject to serious tech nical uncertainties associated with measurement error, in forms that the data do not support. We had hoped to see these issues discussed as well, again from an international perspective. Other subjects on which we would have liked additional papers relate to statistical activities in developing countries and the transfer of technology. Despite our best efforts, authors from developing countries are underrepresented in this volume. Some reasons for this may be gleaned from the paper by Pearce. We also failed in our efforts to obtain a paper on the transfer of technology from one developed country to another. We had in mind the Japanese effort in the area of quality control and its transfer to the factories of older industrialized nations. Finally, in our initial round of invitations we attempted to secure papers that reflected the varied areas of application of present-day statistical methodology. As we review the list of papers actually written, we note that we have had some success in that a large fraction do deal with applications. Yet there are missing subject matter areas, three notable ones being: business and management, en gineering, and environmental science. There are also methodological topics that have not received much attention. We can only note that the absence of an extended discussion of your favorite statistical method is unintentional. It may seem strange that in a centenary volume there is so little explicit mention of the history of the International Statistical Institute. One reason for this is that the first 75 years of the Institute are chronicled by J.W. Nixon in A History of the International Statistical Institute, 1885-1960, published for the Preface ix 75th anniversary in 1960. But, in addition, we and our advisors decided that it would be more appropriate, and more fun for our readers, if instead of a history we were to produce a volume on the "stuff" of statistics, and to point to the future of our field instead of to its past. All of the papers in this volume have been refereed and many have been thoroughly revised, sometimes more than once. We are grateful to the authors, the members of our advisory panel, and the referees for all of their work and cooperation. The result of their efforts is a much improved collection of papers. We believe that there is something here for anyone who is interested in, or curious about statistics. We hope that you agree, and that you take from this volume a sense of the excitement with which we view the future of statistics. ANTHONY C. ATKINSON STEPHEN E. FIENBERG Acknowledgements We are indebted to a large number of people who assisted in the preparation of this volume. David Cox served as liaison with the lSI, and we turned to him for advice and assistance on numerous occasions. Our Editorial Advisory Panel provided suggestions for authors and topics as we began to get organized, and the members served as primary reviewers of most of the manuscripts. A.L. Finkner gave us special assistance and suggestions in the area of sample surveys. In addition, the following individuals served as referees of one or more papers: R. A. Bailey, O. E. Bamdorff-Nielsen, D. J. Bartholomew, J. L. Bell, J. T. Bonnen, C. Cress, B. S. Everitt, V. V. Fedorov, I. Ford, J. Greenhouse, A. M. Herzberg, Wm. Kruskal, N. Laird, J. Lehoczky, R. Mead, M. M. Meyer, H. D. Patterson, J. N. K. Rao, R. L. Smith, C. C. Spicer, J. M. Tanur, E. J. Thompson, and F. Vogel. Finally, we would like to thank Margie Krest for keeping track of everything as work on the volume progressed, and for her assistance in putting the final manuscript together for the publisher. Contents Foreword by J. DURBIN v Preface Vll Acknowledgements x 1. Prediction and Entropy 1 HIROTUGU AKAIKE 2. Statistical Developments in World War II: An International Perspective 25 STEPHEN E. FIENBERG 3. Statistics in the United Kingdom, 1939-45 31 G. A. BARNARD and R. L. PLACKETT 4. The Fascination of Sand 57 OLE E. BARNDORFF-NIELSEN, PREBEN BLlESILD, JENS LEDET JENSEN, and MICHAEL S0RENSON 5. International Trends in Official Statistics 89 PETTER JAKOB BJERVE 6. Cohort Analysis in Epidemiology 109 N. E. BRESLOW 7. Psychiatric Statistics 145 MORRIS H. DEGROOT and JUAN E. MEZZICH 8. Mathematical Models for Infectious Disease Statistics 167 K. DIETZ and D. SCHENZLE 9. Evolutionary Origins of Statisticians and Statistics 205 J. DURBIN 10. Statistical Computing: What's Past Is Prologue 233 WILLIAM F. EDDY and JAMES E. GENTLE 11. Conditional Econometric Modeling: An Application to New House Prices in the United Kingdom 251 NEIL R. ERICSSON and DAVID F. HENDRY 12. Large-Scale Social Experimentation in the United States 287 STEPHEN E. FIENBERG, BURTON SINGER, and JUDITH M. T ANUR xii Contents 13. The Development of Sample Surveys of Finite Populations 327 MORRIS H. HANSEN, TORE DALENIUS, and BENJAMIN J. TEPPING 14. On Some New Probabilistic Developments of Significance to Statistics: Martingales, Long Range Dependence, Fractals, and Random Fields 355 C. C. HEYDE 15. Concepts of Robustness 369 CHRISTOPHER FIELD 16. Robust Estimators of Location and Their Second-Order Asymptotic Relations 377 JANA JUREtKOVA 17. Mathematical Statistics in the Humanities, and Some Related Problems in Astronomy 393 DAVID G. KENDALL 18. Reflections on the World Fertility Survey 409 MILO~ MACURA and JOHN CLELAND 19. Economic and Social Statistics for Comparative Assessments 437 E. MALINVAUD 20. An Introduction to the Optimum Design of Experiments 465 ANTHONY C. ATKINSON 21. Experimental Design in Russian Practice 475 V. V. NALIMOV, T. I. GOLIKOVA, and Yu. V. GRANOVSKY 22. Agricultural Experimentation in a Developing Country 497 S. C. PEARCE 23. On the Statistical Analysis of Floods 511 LUIS RAUL PERICCHI and IGNACIO RODlUGUEZ-ITURBE 24. Weighted Distributions Arising Out of Methods of Ascertainment: What Population Does a Sample Represent? 543 C. RADHAKRISHNA RAo 25. Statistical Problems in Crop Forecasting 571 D. SINGH and M. P. JHA Author Index 585 Subject Index 600