Inferential Errors in Repeated-Measures ANOVAs
Many researchers discuss the results of their fully repeated measures ANOVAs in such a way that it appears they have discovered ultimate Truth. Stated differently, the language used in many research reports suggests strongly that sample statistics and the results of inferential tests are being reified into population parameters and indisputable claims.
You must remember that the result of any F-test might be a Type I error (if the null hypothesis is rejected) or a Type II error (if the null hypothesis is retained). This is true even if the nine-step version of hypothesis testing is used, and even if attention is paid to all relevant underlying assumptions, and even if the data are collected in an unbiased fashion with valid and reliable measuring instruments from probability samples characterized by zero attrition, and even if all other good things are done so the study is sound. Simply stated, inferential error is always possible whenever a null hypothesis is tested.
(From Chapter 14 in the 6th edition, p. 341)
Copyright © 2012
Schuyler W. Huck