Are ANCOVA Studies Better Than ANOVA Studies?
My final warning [in Chapter 15] has to do with your general opinion of ANCOVA-based studies as compared with ANOVA-based studies. Because ANCOVA is more complex (due to the involvement of a larger number of variables and assumptions), many consumers of the research literature hold the opinion that data-based claims are more trustworthy when they are based upon ANCOVA rather than ANOVA. We strongly encourage you to refrain from adopting this unjustified point of view.
Although ANCOVA (as compared with ANOVA) does, in fact, involve more complexities in terms of what is involved both on and beneath the surface, it is an extremely delicate instrument. To provide meaningful results, ANCOVA must be used very carefully--with attention paid to important assumptions, with focus directed at the appropriate set of sample means, and with concern over the correct way to draw inferences from ANCOVA's F-tests. Because of its complexity, ANCOVA affords its users more opportunities to make mistakes than does ANOVA.
If used skillfully, ANCOVA can be of great assistance to applied researchers. If not used skillfully, however, ANCOVA can be dangerous. Unfortunately, many people to think of complexity as being an inherent virtue. In statistics, that is often not the case. As pointed out earlier in the chapter, the interpretation of ANCOVA F-tests is problematic whenever the groups being compared have been formed in a nonrandom fashion--and this statement holds true even if (1) multiple covariate variables are involved, and (2) full attention is directed to all underlying assumptions. In contrast, it would be much easier to interpret the results generated by the application of ANOVA to the data provided by subjects who have been randomly assigned to comparison groups. Care is required, of course, whenever you attempt to interpret the outcome of any inferential test. My point is simply that ANCOVA, because of its complexity as compared to ANOVA, demands a higher--not lower--level of care on your part when you encounter its results.
(From Chapter 15, pp. 403-404)
Copyright © 2012
Schuyler W. Huck