Dear SPMers,
A recent discussion about how best to set up individual differences analyses (e.g. to examine the association between individual neural activation profiles and a person-specific parametric performance index) resulted in various ways of implementing & interpreting parametric modulation analyses being suggested. Some of these suggestions however came with a few question marks, so it would be great to get your thoughts on the following issues.
1.) Is it at all reasonable / valid to run a 1st level analysis on a specific contrast (e.g. stimulus A – stimulus B) and then use the resulting con.img files from every person in a multiple regression analysis (whole brain), so that the betas contained in the con.img files are regressed on a parametric variable (e.g. performance index from a behavioural task) and controlling for demographics (age & gender)?
On a related, yet slightly separate note:
2.) Assuming the respective analyses have been conducted properly, is it possible to obtain no significant activation changes (whole brain analysis, FWE-corrected, p=0.05, no extent threshold) on the group level for a specific stimulus A but to obtain significant clusters (whole brain analysis, FWE-corrected, p=0.05, no extent threshold) when performing an analysis as outlined above (i.e., use the 1st level con.img files for stimulus A in a multiple regression analysis, regressing stimulus A-related betas on a parametric regressor whilst controlling for age & gender)? If it is indeed possible to obtain such results, would these indicate that, overall, the stimulus results in activation changes that are rather small and/or varied, hence there is no ‘mean signal’ for the stimulus itself, but that whatever stimulus-specific signal change exists on the individual level is associated with performance?
Any comments, thoughts or references on this matter would be greatly appreciated!
Thank you & kind regards,
Loreen
|