Braxton Alfred M. Elements of Statistics for the Life and Social Sciences

Springer, 1987. — 202 p.This book was written to myself at about the time I began graduate studies in anthropology-the sort of thing a Samuel Beckett character might do. It is about the conduct of research. In a very real sense the purpose is partially to compensate for the inadequacies of my professors. Perhaps this is what education is about. The effort has not been an unqualified success, but it has been extremely gratifying.I was trained in anthropology. After completing the Ph.D. I went to Stanford on a post-doctoral fellowship. At the time, this was a novelty and the department was not prepared for such a thing. To stay occupied I began attending lectures, seminars, and discussion groups in mathematics and statistics. This was about the luckiest choice I ever made. The excitement was easily as intense as that which I experienced upon encountering anthropology. On one occasion I innocently and independently proved a theorem that had first been done 2000 years earlier. It is currently used as an exercise in high school mathematics so it is neither difficult nor arcane. Learning all this did not tarnish my sense of discovery. (On reflection I am puzzled by my failure to have seen all this "beauty" when I was exposed to it as an undergraduate. The unparalleled excellence of the Stanford program was undoubtedly responsible for my belated conversion.)Introduction
Some Elementary Principles of Deductive Argument
Common Connectives for Statements
Argument
The Logic of Scientific Argument
The Program of Science
Elements of a Good Test
Examples
Causation, Mill's Methods
Description, Pre-Science, ScienceGenerating Predictions
Introduction and Orientation
Background
Processes
Topics in Hypothesis TestingTesting a One-Dimensional Hypothesis
Testing a Two-Dimensional Hypothesis
Tests of Hypotheses in Three or More Dimensions
Special Topics
Summary and Conclusions
Appendixes
Matrix Manipulation
Conversion of the Base of Logarithms
Bayes'Theorem
Table of Chi-Square Distribution, 5% Points
The Choice of Computing Software for Log Linear Models