An Accidental Statistician Read online

  Table of Contents

  Title Page

  Copyright

  Dedication

  Foreword

  Second Foreword

  Preface

  Acknowledgments

  From The Publisher

  Box Titles Published with Wiley

  Chapter One: Early Years

  Chapter Two: Army Life

  The Experimental Station at Porton Down

  Chapter Three: ICI and the Statistical Methods Panel

  Chapter Four: George Barnard

  Chapter Five: An Invitation to the United States

  Chapter Six: Princeton

  Chapter Seven: A New Life in Madison

  Chapter Eight: Time Series

  Chapter Nine: George Tiao and the Bayes Book

  Chapter Ten: Growing Up (Helen and Harry)

  Chapter Eleven: Fisher—Father and Son

  Chapter Twelve: Bill Hunter and Some Ideas on Experimental Design

  Chapter Thirteen: The Quality Movement

  Chapter Fourteen: Adventures with Claire

  Chapter Fifteen: The Many Sides of Mac

  Chapter Sixteen: Life in England

  Chapter Seventeen: Journeys to Scandinavia

  Chapter Eighteen: A Second Home in Spain

  Chapter Nineteen: The Royal Society of London

  Chapter Twenty: Conclusion

  Chapter Twenty One: Memories

  Index

  Copyright © 2013 by John Wiley & Sons, Inc. All rights reserved

  Published by John Wiley & Sons, Inc., Hoboken, New Jersey

  Published simultaneously in Canada

  No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning, or otherwise, except as permitted under Section 107 or 108 of the 1976 United States Copyright Act, without either the prior written permission of the Publisher, or authorization through payment of the appropriate per-copy fee to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, (978) 750-8400, fax (978) 750-4470, or on the web at www.copyright.com. Requests to the Publisher for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, (201) 748-6011, fax (201) 748-6008, or online at http://www.wiley.com/go/permission.

  Limit of Liability/Disclaimer of Warranty: While the publisher and author have used their best efforts in preparing this book, they make no representations or warranties with respect to the accuracy or completeness of the contents of this book and specifically disclaim any implied warranties of merchantability or fitness for a particular purpose. No warranty may be created or extended by sales representatives or written sales materials. The advice and strategies contained herein may not be suitable for your situation. You should consult with a professional where appropriate. Neither the publisher nor author shall be liable for any loss of profit or any other commercial damages, including but not limited to special, incidental, consequential, or other damages.

  For general information on our other products and services or for technical support, please contact our Customer Care Department within the United States at (800) 762-2974, outside the United States at (317) 572-3993 or fax (317) 572-4002.

  Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic formats. For more information about Wiley products, visit our web site at www.wiley.com.

  Library of Congress Cataloging-in-Publication Data:

  Box, George E. P.

  An accidental statistician : the life and memories of George E.P. Box / George E. P. Box.

  pages cm

  Includes index.

  ISBN 978-1-118-40088-3 (cloth)

  1. Box, George E. P. 2. Statisticiansendash United Statesendash Biography. I. Title.

  QA276.157.B69A3 2013

  519.5092–dc23

  [B]

  2012040251

  This book is dedicated to my students, with whom

  it was my privilege to work, and who became my friends.

  Foreword

  Virginia Woolf wrote about a character with a mind that “kept throwing up from its depths, scenes, and names, and sayings, and memories and ideas, like a fountain spurting over.” George Box is the embodiment of that active mind. Dinner with George is a spurting of stories, poems, songs, and anecdotes about his work and his friends. An Accidental Statistician jumps you into that fountain of ideas. It is great fun even if books about statistics and science are normally absent from your reading list.

  No doubt many readers think, as I once did, that the subject is difficult and dull. Here we have a charming and colorful storyteller who quotes Yogi Berra in a discussion on the analysis of variance; has Murphy, of Murphy's Law fame, ringing the alarm when there is an opportunity to make things better; and explains an experiment with critical variables named “banging” and “gooeyness.” There are stories about composite designs, time-series forecasting, Evolutionary Operation, intervention analysis, and so on, but they are not mathematical, and most include personal anecdotes about people who were involved in their invention and original application. You learn about statistics and science and, simultaneously, meet a literal Who's Who of statisticians and scientists, and the Queen of England as well.

  I met George Box in 1968 at the long-running hit show that he called “The Monday Night Beer Session,” an informal discussion group that met in the basement of his house. I was taking Bill Hunter's course in nonlinear model building. Bill suggested that I should go and talk about some research we were doing. The idea of discussing a modeling problem with the renowned Professor Box was unsettling. Bill said it would be good because George liked engineers. Bill and several of the Monday Nighters were chemical engineers, and George's early partnership with Olaf Hougen, then Chair of Chemical Engineering at Wisconsin, was a creative force in the early days of the newly formed Statistics Department. I tightened my belt and dropped in one night, sitting in the back and wondering whether I dared take a beer (Fauerbach brand, an appropriate choice for doing statistics because no two cases were alike). I attended a great many sessions over almost 30 years, during which hundreds of Monday Nighters got to watch George execute an exquisite interplay of questions, quick tutorials, practical suggestions, and encouragement for anyone who had a problem and wanted to use statistics. No problem was too small, and no problem was too difficult. The output from George was always helpful and friendly advice, never discouragement. Week after week we observed the cycle of discovery and iterative experimentation. We saw real examples that, “All models are wrong, but some are useful.” We saw how statistics is a catalyst for scientific method, and how scientific problems catalyze ideas for doing statistics. What a treat.

  My business is water quality engineering. One night I wanted to discuss a problem that involved a measurement called the Biochemical Oxygen Demand (BOD). George asked whether it would be all right for him to explain this BOD test. He gave as good an explanation as I ever heard. I asked how he happened to know that and learned that at age 16, he took a job as an assistant chemist in a sewage treatment plant. One year before I was born, in 1939, at age 19, he published a paper about oxygen demand in the activated sludge wastewater treatment process, which at the time was new and poorly understood. George's paper can stand with papers on the subject written by some famous Wisconsin engineers who worked on the problem at about the same time. In the 1990s, 55 years later, George and I worked on forecasting the dynamics of activated sludge process performance using multivariate nonstationary time series. Imagine that from a world-famous statistician who was one of the ear
liest researchers on this widely used wastewater treatment process.

  George and I have one bit of unfinished piece of business. A 20-foot-tall civil war soldier guards the stone arch entrance to Fort Randall Park, which is next to the engineering building and the statistics building at the University of Wisconsin—Madison. George had thought for some time that the soldier should have a medal. In 1993, we found a suitable brass medal in a sidewalk stall near Hyde Park, London, but our plan to hang it around the soldier's neck was never fulfilled. I now believe that the medal should stay with George as an award of merit for memoir writing. He deserves it. (And we two old friends do not have to climb the soldier.)

  Last night, May 10, 2012, my wife and I had dinner with George and Claire at their house in Shorewood Hills. He said, “The memoir is finished.” I asked, “What's your next project. It's hard to picture you not doing some writing every day.” He answered, “I 'm thinking of a paper about Fisher's idea on multiplicative effects in experiments.”

  An Accidental Statistician is finished, but apparently George is not. That is good news. Thank you, George.

  P. Mac Berthouex

  Emeritus Professor

  University of Wisconsin

  Second Foreword

  It is a pleasure to welcome this autobiography of an extraordinary scholar and gentleman: George Edward Pelham Box. My intersection with George Box's long and active statistics career begins in the summer of 1952, when I was a research associate at the Aberdeen Proving Grounds, MD, working for Dr. Frank Grubbs. Dr. Grubbs and I together read George Box's wonderful paper on response surface experimentation,1 and I recall how we both marveled at its simplicity … obvious of course now that the message was offered! Originality and elucidation are the signs of Box's genius. On returning that September to my graduate studies at the Institute of Statistics, NC State, in Raleigh, I learned that George Box had agreed to come to Raleigh for a year as a visiting Research Professor. In January 1953, I became his first graduate student.

  It's a pleasure to read about George's boyhood. His father worked hard to provide a modest family environment in a society that offered advantaged youths the greatest opportunities. We learn how George, through the good fortune of meeting alert teachers, uncovered his talents as a writer and more modestly in mathematics. He starts his young adult life as an assistant chemist collecting data on wastewater treatment, and it is here that his career as a statistician begins. And what a career it proved to be!

  One of George Box's distinguishing characteristics is that he only occasionally published a paper as a sole author. This memoir introduces us to a host of his students and research associates, and it provides colorful vignettes of these many wonderfully varied personalities who became his co-authors. Some ten years ago, a group of his students gathered together a compendium of his papers covering the statistical fields of quality, experimental design, control, and robustness.2 Both neophytes and savants have found the exposition within these many papers superb. Of course, there are also his co-authored textbooks on experimental design,3 time series analysis,4 Bayesian inference,5 and control,6 wherein elucidation of the subject's theory and application repeatedly prove both original and illuminating. For George Box, the acronym “KISS” translates into “Keep It Sophisticatedly Simple.”

  I found one recollection of George Box's early statistical experiences particularly fascinating. Near the end of World War II, the British came into possession of German explosive shells containing unknown deadly gases. George was part of the small group that first determined the spectacular deadliness of tiny concentrations of these new reagents (actually nerve gases). They were never employed in warfare, but had portions been dropped over major cities, the population consequences could easily have rivaled those of Hiroshima and Nagasaki. To have been present working along that thin boundary keeping the world from additional disaster is impressive.

  We learn, too, of those whom George Box considered his mentors and hear tales of his early collaborations with others well known among the statistical fraternity. And throughout this autobiography, we become aware of Box's broad contributions to modern statistical theory and practice. His papers and books have vastly expanded awareness of Bayesian methods and time-series modeling. We find the production of information-laden data to be a statistical specialty that enhances scientific progress as it moves from initial conjecture through experimentation and data analysis, leading on to new conjecture. And beyond statistically formal matters, we also capture his deep appreciation for the mind's ability, or better still that of a collection of minds, to give birth to new models and not usual conjectures. It seems that George Box's advice to all those pondering a problem is to be sure to think out of the box.

  J. Stuart Hunter

  Professor Emeritus

  Princeton University

  1 “On the Experimental Attainment of Optimum Conditions” Box, G. E. P. and Wilson, K. B. (1951) Jour. Royal Stat. Soc. Series B, 13, 1–45.

  2 Box on Quality and Discovery: George Tiao, Søren Bisgaard, William J. Hill, Daniel Peña, Stephen M. Stigler. (2000) John.Wiley & Sons.

  3 Statistics for Experimenters: Box, G.E.P, Hunter, William G. and Hunter, J. Stuart (1978) John Wiley & Sons.

  4 Time Series Analysis: Forecasting and Control: Box, G.E.P and Jenkins, Gwilym M. (1970) Holden Day.

  5 Bayesian Inference in Statistical Analysis: Box, G.E.P. and Tiao, George C. (1973) Addison Wesley.

  6 Statistical Control by Feedback and Adjustment: Box, G.E.P. and Lucño, Alberto (1997) John Wiley & Sons

  Preface

  There is a story about a very tall man who was walking with his four-year-old son to pick up a newspaper. He suddenly realized that his son was having difficulty keeping up with him. He said, “Sorry, Tommy, am I walking too fast?” And the boy said, “No, Daddy, I am.”

  Now this account can be viewed in two ways: as an amusing story, or joke or, as illustrating the essence of scientific discovery. The boy's view of the situation was correct but not obvious. The father's view was obvious but wrong.

  So it is perhaps no coincidence when humor and scientific insight come together. Good science is a form of wit, of seeing the joke that nature is playing on us.

  At 93, I can look back on quite a few examples.

  The Box family circa 1895. From left clockwise, Uncle Bertie, my grandfather and grandmother, my father, Aunt Daisy, Uncle Pelham, and Aunt Lina.

  Acknowledgments

  For much of the time that this memoir was being written, I have been quite ill. This has placed an enormous burden on my wife, Claire, who herself has been ill during much of this time. Had it not been for her devoted help, so generously given, this memoir could never have been written. This help has been realized by a practical, ingenious, and well-trained mind. Whenever there is a crisis, she has not only known what to do, she has done it with cheerful understanding and expertise.

  In addition, I am especially grateful to:

  Bovas Abraham Judy Hunter

  David Bacon Stu Hunter

  Ford Ballentyne Brian Joiner

  Ernesto Barrios Tim Kramer

  Mac Berthouex Kevin Little

  Sue Berthouex Alberto Luceño

  Claire Box Merve Muller

  Joan Box Vijay Nair

  Helen Box Lars-Erik Öller

  Harry Box Judy Pagel

  Robin Chapman Daniel Peña

  Norman Draper José Ramírez

  Conrad Fung Marian Ros

  Larry Haugh Xavier Tort

  Margaret Homewood John Sølve Tyssedal

  Brent Nicastro for his permission to use various photographs.

  And to Judith Allen, my friend.

  From The Publisher

  To those of you who do not know George Edward Pelham Box well, suffice it to say that he is a titan in the field of statistics. He is a self-taught statistician who utilized his experience and knowledge of statistics to create unique contributions to many areas particularly in process improve
ment. And, he is a nice guy, to boot. He rarely—if ever—needs an introduction. His very presence is our present.

  This book is being published in the year of George's ninety-fourth birthday as a memoir of his life, his friends, and his contributions. We know that, during his academic tenure, he wrote over 2000 journal articles; published twelve books for Wiley alone (see list below) resulting in over a quarter-of-a-million copies sold worldwide; and was responsible for helping to get Technometrics, a joint publication of the American Society for Quality and the American Statistical Association, off the ground.

  We also know from first-hand experience that George is a true gentle-man, a loving father, and a dedicated husband. He has influenced the lives, in no small way, of everyone he has touched, from young aspiring statisticians to experienced editors-in-chief. When you see a grin on his face, you know that he is about to espouse a bit of wisdom mixed-in with a tad of advice and always with a joke. He most often accomplishes what he sets out to achieve, without fanfare or accolade. He is probably the most unforgettable character a graduate student or editor-in-chief has ever had the experience and pleasure to know.

  Two people have assisted George in the production of this book. They include his loving wife of twenty-seven years, Claire Box, and his friend and research assistant, Judith Allen, both of whom supported him as he wrote this book.

  The management and staff of Wiley commend Dr. Box for all that he has done to enrich the world of statistics, both here and abroad. We wish him continued “presence” and the peace of mind that he will always remain a titan in the written word and in our hearts for generations to come.

  Box Titles Published with Wiley

  Evolutionary Operation: A Statistical Method for Process Improvement Box-Draper, 1969

  Statistics for Experimenters: An Introduction to Design, Data Analysis, and Model Building Box-Hunter-Hunter, 1978

  Empirical Model-Building and Response Surfaces Box-Draper, 1986