 
Summary: 1
R.A.Fisher on the Design of Experiments and Statistical Estimation
(Presented at the Center for History and Philosophy of Science, Boston Univ., March 1990.)
0. Introduction and outline.
On this occasion of the centenary year of Fisher's birth, my purpose in this talk is to consider the
relation between two of Fisher's major contributions: his theory of experimental design and his
theory of statistical estimation. It is no coincidence that each of Fisher's three, principal books on
statistics ends with a substantial chapter on statistical estimation:
Statistical Methods for Research Workers [SMfRW] 1925  14th ed. 1970;
The Design of Experiments [DoE] 1935 8th ed. 1966;
Statistical Methods and Scientific Inference [SM&SI] 1956  3rd ed. 1973.1
The thesis of my presentation is that Fisher linked experimental design and estimation through
his technical account of (Fisher) Information. In particular, improvements in an experimental
design, e.g., better controls, blocking, or other factorial restrictions, may be quantified by an
increase in the Information provided by estimates derived from the experimental data.
In section 1 I sketch Fisher's theory of estimation with an eye on explicating the role Information
plays in resolving choices of rival estimates derived from a sample. As an illustration of this
approach, I show how Fisher's 2
significance test for ordinary contingency tables may be
decomposed to reveal that Information justifies maximum likelihood estimation. The same
