Dear Sir,
Thank you very much. Just to clarify the matter I would like to comment your
letter and also ask one follow up question
>If you're interested in doing an ROI analysis, you'll still probably
>want to use the GLM framework to create a first level analysis of
>your data.
Yes of course, I already did that
>You can either do this as normal and then mask the
>results with your a priori ROI, or you could possibly replace the
>brain mask that feat generates with your ROI mask and only analyse
>that portion of the signal you collected. The two are effectively
>equivalent, but the specifics on how to do multiple comparison
>thresholding vary slightly in practice. I would recommend the former for ease.
I understand this, I am using the “usual” that is the first approach
>If you expect that within your ROI, all of the signal (ie all voxels)
>should all be related to your conditions, then there is no multiple
>comparison problem because you are treating the region as a single
>measurement. in that case, you can use featquery to get a mean %
>signal change estimate across conditions and enter those values (per
>subject) into something like SPSS.
Here major concern comes. I assume that you are referring to mean PE change,
however I am not 100% sure – sorry :-(
Best regards
Michal Kuniecki Ph.D.
Dpt. of Psychophysiology
Institute of Psychology
Jagiellonian University
Al. Mickiewicza 3
31-120 Kraków
|