Hi all,
We are running an effort-based decision making paradigm and we had effort level and reward magnitude pmods entered at certain cue conditions. With multiple pmods, orthogonalization would give preference to the first entered pmod so it might give us different solutions for different orders of pmod entry. We were seeing interesting results with our reward pmod, but because that was entered second, we wanted to see how solutions might change is the order were changed. We followed Tor Wagers suggestions (http://wagerlab.colorado.edu/wiki/doku.php/help/fmri_help/fmri_statistical_models/parametric_modulation) and created a few different models (note these models were created primarily to help us understand our results and aid in interpretation). We ended up with 5 models in total, including our original. The first model (the original) had 2 pmods for some cue conditions, where effort was entered first. A second model created redundant cue conditions so that each cue would only have 1 pmod, either effort or reward, to try to avoid the multiple pmod issue. The third model just had reward entered as pmod, without effort at all. The fourth model was the same as the first, but the spm_orth commands were commented out to make SPM not orthogonalize. And finally, the fifth model was the same as the 1st except that reward was entered first, then effort.
The first 5 models yielded the same results. However, the fifth model (where both effort and reward were entered as pmods, with reward first) yielded different results. The results of this model were in the same general regions as the other 4 models but were considerably stronger.
Can someone help me understand why this is the case and what this means/how this might aid in my interpretation of any model with multiple pmods?
Thank you in advance!
|