"as some parameters are derived from all the volumes of a model" - probably not due to the volumes as such, but due to the different design matrices / different whitening / different variance-covariance matrices, no idea. In any case, raw values in xX.X and raw global estimates xGX.gSF still agree well between separate models and the multi-session model, but the filtered whitened values in xX.xKXs.X do not.
|