Hi Alfonso,

Just on the back of this, could you clarify what CONN would be using instead of Wilks' Lambda and how the SPM ReML approach (presumably for dealing with heterogeneity of variance) is implemented for multiple dependant variables? I can't think how this is possible given that you would get different heterogeneity estimates per-DV and so couldn't filter the design matrix in the same fashion (unless you used multiple design matrices)?

Best wishes,
- Martyn