Thank you Cyril and Donald,
I know that including covariates increases the sensitivity of my analysis and I've carefully read the recommendations to mean center the covariates. The power to distinguish between the effect of each covariate was my concern and you clarified this. But I have another concern: although the covariates weren't used to define the groups, they differ significantly between groups. In this situation the between group contrast (1 -1 0 0 ...) will be identical to the covariate effects (0 0 1 0 ...) and the problem of collinearity remains. Did I get this right?
Best regards,
Paulo Marques.
|