GLM F F results from subset of contrasts images too optimistic?

36 views

Skip to first unread message

Adam Teed

unread,

Oct 12, 2018, 4:32:21 PM10/12/18

to Aaron's fMRI matlab tools.

Hello Aaron,

I have a 2x5 mixed design (Drug_con*stimulus + random(subID|stimulus)) for which I am doing supplemental analyses of a subset of my datato account for questionnaire composite scores as a covariate. The issue is that each of the two scores does not apply to 2 of levels of our 5 level stimuli variable (photo of person A, matched photo A, person B, matched photo, control photo) and should not be crossed over. For that reason I am modeling in GLM Flex Fast 4 a subset of the contrast images from the 1st level glm that the scores apply to, but the results are too good to be true, regardless of whether I actually include the covariate (Drug_con*stimulus+CF_Mom1 + random(subID|stimulus)), with much more extensive activation and t-scores that are double that of the full model. Considering that it isn’t uncommon to use a subset of data to look at 3-way interactions in GLM Flex Fast, do you know why this might be the case in this instance? I appreciate any feedback!