we performed a multivariate analysis of covariance (MANCOVA) between two groups using fractional anisotropy (FA), axial diffusivity (AD), radial diffusivity (RD), 

and mean diffusivity (MD) of the inferior fronto-occipital fasciculus as dependent variables. 

Because the two groups differ also in global (across the whole white matter) FA, AD, RD and MD values, we used all these four measures simultaneously as covariates

of no interest.

However, some authors challenge the usage of all four global measures simultaneously as covariates of no interest and denoted this procedure an inappropriate “overcorrection”.

Is this procedure statistically problematic?

Is it more appropriate to first regress out the global effect of each single measure and then run the multivariate analysis just on the residuals performing an MANOVA instead of a

MANCOVA model?

Any suggestions is highly appreciated.

Thanks in advance

Best regards

Jürgen Hänggi