We have weird variability in the PPI betas.
I just recently ran a PPI analysis on a spatial working memory task. The task contains 4 separate loads, and 3 delay conditions. For running the PPI, I combined all the onsets for each load into single files. That is there is separate task onset files for Load 1, Load 3, Load 5, Load 7, and All Loads.
We used the fsl motion outliers tool before running first levels. We used the gPPI approach, and set up each onset files, the seed time course, motion parameter files, and PPI for each onset file X time course as EVs in the first levels. For the group analysis, we set it up to show group PPI for patients and controls for all the PPI files, Load 1 PPI, Load 3 PPI, Load 5 PPI, and All Loads PPI.
We extracted betas using the following command:
fslmeants -i filtered_func_data -m mask
Now onto the issue, looking at the betas for each subject in the All Loads PPI (we originally wanted to correlate the values with a behavioral measure) we notice large variability in the beta values between subjects. The values range from -33224.96082 to 21866.67177. The variability within subject seems to be quite large as well. One of our colleagues looked at her betas from a previous PPI (with a different task and different subjects) did not have such large variability in betas, although the data was run with the basic PPI not gPPI.
The variability does not seem to come from the raw data, so it must arise at some point during processing. The variability exists in both patients and controls. Is this something we should be concerned about? Any ideas where the variability comes from?
Any help on this issue would be much appreciated.
Best,
Amira Ibrahim
|