Page 632 - Probability and Statistical Inference
P. 632
14. Appendix 609
Science, to Neyman. This monograph had a life-long influence on Neyman.
Very early in college, Neyman was keenly interested in the theory of Lebesgue
measure. But, life-style was uncertain and difficult during those days of wars.
Poland and Russia started fighting about territories. Because of his Polish
background, Neyman was arrested as an enemy alien and he spent a few
weeks in jail. But, he was needed in the University to teach, and hence he was
ultimately let go.
At the age of 27, Neyman first had the opportunity to visit Poland in an
exchange of prisoners of war and in Warsaw, he met the Polish mathemati-
cian W. Sierpi ski who encouraged him and thought highly of his research.
Ultimately Sierpi ski helped Neyman to get a job as a statistician in the Na-
tional Institute of Agriculture. Here, he was to take meteorological observa-
tions and help with agricultural experiments.
Neymans doctoral thesis (1923) in the University of Warsaw had dealt
with probabilistic considerations in agricultural trials. In 1924, he went to the
University College London for a year to study under Karl Pearson where he
met three other statisticians: R. A. Fisher, E. S. Pearson, and W. S. Gosset
(Student). Some of Neymans papers were known to the people at the
University College because their English translations were already available.
Neyman spent the following year in Paris with a Rockefeller Fellowship, and
became acquainted with H. Lebesgue, J. Hadamard, and E. Borel.
In late 1920s, Neyman focused more upon mathematical statistics as well
as applications in economics, insurance, biology and industry. His collabora-
tion with E. S. Pearson started around 1925. Since early 1920s, (E. S.) Pearson
began developing his own philosophy of statistical methods and inference. He
also started to appreciate and build practical statistical models, particularly
useful for industrial applications. During 1928-1938, extensive collaborations
took place between (E. S.) Pearson and Neyman.
Neyman and (E. S.) Pearson (1928a,b) approached inference problems to
build statistical tools for experimenters to choose between two classes of
models. These culminated later into path-breaking contributions, Neyman and
(E. S.) Pearson (1333a,b). In the latter papers, likelihood ratio tests in the
multi-parameter cases were fully developed. Neyman and Pearson (1933a)
has been included in the Breakthroughs in Statistics, Volume I [Johnson and
Kotz (1992)]. Neyman-Pearsons formulation of optimal tests ultimately evolved
into optimal decision functions for more general statistical problems in the
hands of Wald (1949b,1950).
Fisher criticized the Neyman-Pearson approach claiming that his estima-
tion theory, along with the likelihood and sufficiency, was quite adequate, and
that the work of Neyman and Pearson on testing of hypotheses was mis-
