ANCOVA's Underlying Assumptions

As indicated earlier, the "standard" assumptions of ANOVA (e.g., normality, homogeneity of variance, sphericity) underlie ANCOVA as well. You should upgrade or downgrade your evaluation of a study depending upon the attention given to these assumptions in the situations where F-tests are biased when assumptions are violated. To illustrate this point, consider Excerpt 18.22 [not shown here]. Notice how the researchers tested the equal variance assumption (in addition to the linearity and homogeneity of regression slopes assumptions). As a recipient of research claims, you have a right to expect researchers to be no less concerned about underlying assumptions than was the research team that conducted the study from which Excerpt 18.22 was taken. Unfortunately, you are likely to come across many ANCOVA studies in which there is absolutely no discussion of linearity, equal regression slopes, or homogeneity of variance.

(From Chapter 18, p. 544)

Copyright © 2003 Schuyler W. Huck
All rights reserved.