Repeated measurement: sensitive tests for experiments with few animals.

Abstract

Animal scientists who conduct experiments involving repeated measurements of animals often are frustrated by low statistical power of tests for comparisons of treatment means. In many cases, low power of the traditional tests simply is the consequence of low replication (few animals per treatment) that was forced by cost or complexity of experimental technique. A method is given for comparing treatments in a way that permits sensitive tests (via partition of the treatment X period interaction), often when the number of animals per treatment is not more than five or six. Modifications for the procedure are given for the case of heterogeneous variances and covariances. An example from mammary physiology is used to illustrate the procedure and to compare it with standard methods of analysis.