George W. Snedecor, William G. Cochran.
Statistical Methods: 6th edn.
Iowa State University Press. 1967
(read but not reviewed)
New topics in this edition are:
• Analysis of data recorded in scales having only a small number of distinct values
• Prediction of the independent variable X from the dependent variable Y
• Linear regression when X is subject to error
• Comparison of two correlated estimates of variance
• Analysis of proportions in ordered classifications
• Testing a linear trend in proportions
• Analysis of a set of 2 x 2 contingency tables
• Remedial measures for the effects of failures in the assumptions of the analysis of variance
• Selection of variates for prediction in multiple regression
• The discriminant function
• The general method of fitting non-linear regression equations
and its applications to asymptotic regresssion
• Analysis of proportions in two-way tables with unequal numbers of observations >
Suggestions from teachers who have used past editions have brought about several structural changes in the sixth edition. The material on large sample methods is now presented earlier and an introduction to probability is followed by the binomial and Poisson distributions. The discussion of multiple regression precedes the presentation of covariance and multiple covariarice. The statistical tables have been placed in an Appendix. For a one term course, teachers may use the suggested “Short Course” preceding the Table of Contents.
As in past editions, the mathematical level required involves little more than elementary algebra; dependence on mathematical symbols has been kept to a minimum. Also retained is the extensive use of experimental sampling to familiarize the reader with the basic sampling distributions that underlie modern statistical practice.
While most of the numerical illustrations continue to be from biological investigations, the authors have broadened the range of fields represented by data.
The use of computers and flexible standard programming is fast increasing the investigator’s analytical power in handling data, but, as the authors point out, investigators still need to learn the standard techniques of analysis and understand their meaning in order to check computerized data with full confidence. Statistical Methods will continue to serve these purposes.