|
|
Oscar Kempthorne
1919-2000
|
by Klaus Hinkelmann
Born into a farm family on January 31, 1919, in St. Tudy,
Cornwall, Oscar Kempthorne made up his mind very early that he wanted
to get away from the backbreaking work on the farm. And he realized that
this could only be accomplished through more education, or as he put it:
"There was only one way to do it-brain power" (Des Moines Register,
11/25/90). As a consequence, he studied very hard on his own, but, fortunately,
also some of the teachers in his rural school recognized his talents and
supported him in his quest. He especially loved mathematics, and in secondary
school his knowledge surpassed that of his teachers. So, he taught himself
mathematics and, thus, prepared himself for a university education.
In 1937 his hard work paid off as he won scholarships to
Cambridge University. In his first two years there he took a great number
of mathematics courses, but "got turned off from pure mathematics
because it did not seem to be going anywhere" (Folks, 1995). Also
a brief foray into mathematical physics did not catch his imagination.
Then, finally, he had a first course in statistics from John Wishart,
which seemed to interest him. Inspired by some of R. A. Fisher's writings,
the idea that mathematics should be useful and could be useful in agricultural
research impressed him tremendously. He subsequently took a two-term statistics
course from J. O. Irwin, which would complete his formal training in statistics.
In 1940 he received a B.A. degree with Honours from Cambridge University
and the M.S. degree in 1943.
After working briefly for the British Ministry of Supply,
he joined Rothamsted Experiment Station in 1941. Here he met R. A. Fisher
and Frank Yates. These two men had contributed greatly to the prominence
of Rothamsted through their groundbreaking work in experimental design,
a field to which Kempthorne later also would make important contributions.
During his time at Rothamsted, Kempthorne had actually very little contact
with Fisher. As Kempthorne put it himself "He was there, but I was
a boy; I didn't know enough to ask him a question or to speak to him about
anything" (Folks, 1995). Yet, he regarded Fisher as having the greatest
influence in his professional life and he thought that Fisher was, by
far, the greatest statistician that humanity has produced (Folks, 1995).
It is, therefore, not surprising that Kempthorne became attracted to the
areas of statistics that had been shaped in big measure by Fisher: experimental
design, genetic statistics, and statistical inference.
Kempthorne left Rothamsted when he was asked to participate,
as a member of a technical statistical group, to observe and evaluate
the Greek elections in 1946. This work brought him in contact with Ray
Jessen, a survey sampling expert from Iowa State College and the leader
of the group. This was at the time that a Statistics Department was being
established at Iowa State College. Following the Greek mission, he was
offered a position as Associate Professor at Iowa State in 1947. This
was the beginning of an association that would last 42 years until his
retirement in 1989. He was promoted to Full Professor in 1951 and was
named Distinguished Professor in Sciences and Humanities in 1964.
The Statistical Laboratory at Iowa State had achieved great
prominence through the work of George W. Snedecor and the staff he had
assembled between 1933 and 1947 (for a detailed account see David, 1984).
The establishment of the Statistics Department would add greatly to the
statistical activities at Iowa State, and Kempthorne would come to play
an important role in expanding and promoting these activities through
his teaching, research, writing, and consulting.
His first great success came with the publication of his book Design and
Analysis of Experiments published by Wiley in 1952. It was the first comprehensive
treatment of experimental design at a "theoretical" level intended
for graduate studies and research in this field. It built upon the foundations
laid by Fisher and Yates, but Kempthorne was able to carry the development
much further and thus paint a much broader and unified picture of this
field. Fisher's notion of randomization was cast in a more mathematical-probabilistic
framework, which led to important consequences with regard to the development
of linear models and modes of inference resulting from the analysis of
variance. Based on randomization theory, he pointed out the asymmetry
of blocking and treatment factors in the linear model in that one can
test hypotheses about treatment factors, but not about blocking factors
(these ideas were exposited more fully in the 1994 book Design and Analysis
of Experiments. Vol. 1. Introduction to Experimental Design by Hinkelmann
and Kempthorne). A comment by Folks (1984) puts this idea in the right
perspective: "As a beginning doctoral student at Iowa State University,
I made the comment to him [Kempthorne] that a two-factor experiment and
a randomized block design resulted in the same model. I learned very quickly
that my education was not complete-that there was a fundamental difference
between these two situations resulting from the randomization". In
addition to his pioneering work on randomization, Kempthorne provided
a unified treatment of the general sn factorial experiment with s being
a prime or prime power. With a new parametrization for the observations
from such an experiment, he laid a more mathematical framework for and
brought new insights to the important concepts of systems of confounding
and fractional factorials.
Amazingly, only five years later Kempthorne published,
again with Wiley, another groundbreaking book entitled An Introduction
to Genetic Statistics. In the preface he writes: "My basic debt with
regard to the whole book is to Sir Ronald A. Fisher. There is hardly a
page, even less a chapter, which does not contain the results of Fisher
and extensions of Fisher's results" (Kempthorne, 1957). It also makes
use of important contributions due to Sewall Wright, J. B. S. Haldane,
and G. Malécot, but it is the extensions and applications of the
results of these and other writers that make this book such an important
addition to the genetics and statistics literature. Many of these extensions
and new results represent Kempthorne's own research, which was heavily
stimulated by the environment in which he found himself at Iowa State
and the interaction he had with plant and animal breeders, among them
Jay L. Lush, one of the premier animal breeders at the time. Although,
as he points out: "The present book is not a book on breeding theory
and procedures", it nevertheless provides the theoretical framework
and underpinnings for such procedures with its detailed development and
discussion of fundamental concepts in the context of both population genetics
and quantitative inheritance. Among these concepts are selection theory,
the theory of inbreeding, the notion of covariances between relatives
for various types of genetic populations and mating types within populations,
as well as the role of analysis of variance as it applies to quantitative
inheritance and the estimation of genetic parameters, such as genetic
variance components and combining abilities. It is no wonder that, with
such a broad exposition, this book became a standard text and provided
the tools for many developments in quantitative genetics research today.
A third area to which Kempthorne contributed heavily, in
particular in the latter part of his professional life, is that of the
foundations and philosophy of statistical science. It is interesting to
note that as a student he "did not really understand what philosophy
was about or that there were such people as philosophers" to which
he adds " and I am not much better off now about that" (Folks,
1995). That, of course, is a typical Kempthorne understatement. He became
an avid reader of the works of many of the important philosophers, like
Kant, Schopenhauer, Nietzsche (as somebody who abhors philosophy I remember
with some embarrassment my occasional failures to properly explain to
him the meaning of some of the German words used by these philosophers)
or philosophers of science, like Carnap, Peirce, Popper. He thought and
wrote a great deal about how their ideas applied to the field of statistics
and its foundations. Much of his own philosophy can be found in the book
Probability, Statistics, and Data Analysis (Kempthorne and Folks, 1971),
published by Iowa State University Press. He drew a sharp distinction
between theoretical and applied statistics and felt that data analysis
was at the heart of statistics. He instilled this view in his students
and in the overall direction of the Statistics Department at Iowa State
when he says: "What accounts for the success of the Stat Lab [at
Iowa State]? I believe that it is because it was not driven by the mathematics,
but by actual problems in biology, genetics, demography, economics, philosophy
and so on. To be sure, the real problems give rise to abstract problems
in statistical inference, which have a fascination of their own. However,
for statistics to remain viable, statistical problems should have their
genesis in real, data-related problems" (Folks, 1995).
For all his broad and varied contributions Oscar Kempthorne
was recognized by his peers and by the scientific community with numerous
awards and honors, such as Fellow of the American Statistical Association,
Fellow and President of the Institute of Mathematical Statistics, Fellow
of the American Association for the Advancement of Science, Elected Member
of the International Statistical Institute, Sc.D. degree from Cambridge
University, Honorary Fellow of the Royal Statistical Society, and an Honorary
Doctorate from the University of Ioannina (Greece).
Kempthorne was a consummate teacher. He literally bombarded
his students with handouts to supplement his lectures and expected them
to work through every word and formula. He loved and was loved by (most
of) his students and feared by them at the same time, because he demanded
a lot. In a "Letter to Friends" on the occasion of his 65th
birthday he wrote: "Give me arrogant but competent young people".
During his 42 years at Iowa State University he directed the work of 43
PhD students, and he always took great pride in their later accomplishments.
In the same "Letter to Friends" he wrote: "I
view myself as a person who is kindly, but sometimes harsh. My wife [Valda
Marie, who was his strongest supporter] tells me so, and correctly. She
says: 'Go off and pursue your holy grail, and forget about the humans
you should pay attention to.' I find myself with friends, all over the
world. How did this happen to a Cornish farm boy? Was it genes, environment
or altruism? I do not know." Surely, those of us who had the privilege
of knowing Oscar Kempthorne, have an answer, and, it is hoped those who
only read about him and read his work, will share in our admiration for
him.
References
David, H. A. (1984). The Iowa State Statistical Laboratory: Antecedents
and early years. In: Statistics-An Appraisal (H. A. David and H. T. David,
eds.) 3-18. Iowa State Univ. Press, Ames.
Folks, J. L. (1984). Use of Randomization in Experimental Research. In:
Experimental Design, Statistical Models, and Genetic Statistics-Essays
in Honor of Oscar Kempthorne (K. Hinkelmann, ed.) 17-32. Dekker, New York.
Folks, J. L. (1995). A Conversation with Oscar Kempthorne. Statist. Science
10 321-336
