covariates effects in an ANOVA

#1
Is it strange if I run a repeated measure ANOVA with say 2 covariates, get some results, then run it again with another covariate added (i.e. 3 covariates), and get different results?

This strikes me as odd... I don't see how adding an additional covariate to the analysis should change the other covariates' effects on what's being measured...

This is a repeated measure ANOVA by the way, though I doubt it matters.
 

Dr.D

New Member
#3
There are several assumptions of ANCOVA, two main ones including:

1) A significant relationship between dependent variable and covariate
2) Homogeneity of regression slopes for one covariate - regression planes should be parallel for two covariates.

Also if one of the covariates is highly related with independent variable, this may cause a problem. You run a separate ANOVA where the independent variable is used, and the covariate is treated as dependent variable to see whether this is a significant effect (which is not what you want).

See http://carbon.cudenver.edu/~lsherry/rem/ancova.html for a discussion on this.

Violations of any of these assumptions may be responsible for your findings