Dear Ariadna,
On Sun, 4 Aug 2019 09:39:53 +0100, Ariadna Albajara Saenz <[log in to unmask]> wrote:
>Dear CAT12 experts,
>
>I am comparing three groups using a one-way ANOVA with TIV and age as covariates. Since there was collinearity with TIV, I used global scaling with TIV. Nevertheless, there's still collinearity for age.
>1.What would be recommended in this case? Maybe using age as covariate with interaction and mean centering with factor 1 (group)?
>I hereby attach the design matrix.
That does not look critical. Unfortunately, no threshold or rule of thumb can be given. The higher the co-linearity the more difficult it is to do any correction (except for the TIV which can be then used with global scaling). In case that age would covary this would be difficult to correct for because if you remove variance explained by age you also remove variance that exists between your groups. However, in your case it looks fine.
>2. Also, is the absolute masking threshold still 0.1? (I have also entered the mean TIV in the "Grand mean scaled value").
Yes, if you have entered mean TIV as described.
Best,
Christian
>
>Thank you in advance,
>
>Ariadna
>
>
|