I read that when using ANCOVA to control a co-variate, the covariate and the independent variable should be independent. How about if the covariate DOES vary with the independent variable? In that case what test can I use to adjust for the co-variate?