I'm looking to implement a multiple regression analysis of stimulus ratings (along multiple dimensions) at the 1st level. I've trawled through the list and found only posts referring to 2nd level analyses. My understanding is I can simply run parametric modulation with multiple modulators. However some things remain unclear:
Given that some of the variables are correlated, how would I go about calculating the r^2 for the full regression model? I'm familiar with converting t-stats to r-stats in simple regression models with SPM, but not entirely sure if the same conversion is valid with parametric modulation (don't see why not).
I'm currently thinking that to best approximate vanilla multiple regression in a standard stat package I would:
-Turn off within trial orthogonalization in spm_fMRI_design.m (I want simultaneous entry)
-Enter all 4 stimulus ratings as parametric modulators
-Do separate contrasts for each (e.g. 0 1 0 0 0; 0 0 1 0 0; etc.) and do a T to R transformation on the t-maps which I can then use to calculate r^2 for each predictor.
-For the full model I'd enter a contrast of [0 1 1 1 1] and do a T to R transformation on the t-map to calculate the r^2.