Dear SPM users,
I already posted this question some time ago and haven't received an answer yet. Please excuse if I ask again:
I am currently running an analysis with multiple parametric modulators in SPM12. Since I am interested in the unique effects of each parametric modulator and, therefore, don't want them to be orthogonalized with respect to earlier modulators (as implemented in the default mode of SPM8), I used SPM12 in which I have the opportunity to de-activate the orthogonalization procedure. My question now is: How do I prepare my variables for SPM12? I thought I have to de-mean the variables (which means I center the variables, right?) myself prior to entering them in my design - similar to SPM8 when commenting out the serial orthogonalisation of parametric modulators in the source code. Having done this, however, the main event regressor contrast (against baseline) seems a bit "reduced" in activation clusters compared to my SPM8 version. So the question remains whether I standardize the variables (e.g., by means of z-transforming) in SPM12 in addition to mean-centering them in order to get the right results? At least this variant looks better to me, but I am still not sure if that's the right way to do.
Many thanks in advance,
Sarah
|