Hi all
I have question about multicollinearity issues.
I have data from 10 subjects and 3 runs each. There are 12 regressors (two types of trials x hit or misses x 3 stages in a trial). 1dplot.py on the xmat design matrix tells me that the last two stages of each trial are correlated (>0.7). I used 3dDeconvolve.
If I was interested in closely examining the contributions of these two regressors (last two stages of the trials) at certain brain areas more closely, am I right in following what Erdinez et al did (see attached)?
i.e...
I can pull out my two regressors in question, orthogonalize one with respect to the other in matlab, for instance (with function qr or gram-schmidt orthogonalization) and then use them in GLM models like he/she did to study the contributions of these regressors to common areas in the brain?
OR
Is there another straightforward way to go about this?
Thank you!
Sobana