Hi,
I'm running a group analysis (adults=16, adolescents=11, children=17) third level using outlier detection with 6 EVs (adults, adolescents, children, adult covariate, adolescent covariate, child covariate). One of my contrasts across the first three EVs is [1, -2, 1] specifying a quadratic effect. I find activation for this contrast image in left putamen suggesting that my second group has decreased activation compared to groups 1 and 3. However, when I feat query these data at the first level, I find that my second group has greater activation compared to groups 1 & 3. Examining the copes for EV1, EV2, and EV3, independently, shows a similar finding: greater COPE values for group 2 than groups 1 & 3. However, the zstat images for EV1, EV2, and EV3, show a pattern similar to one I specified in my quadratic contrast: group 2 more negative zstat values than groups 1 & 3.
My first question relates to the finding that my prob_outlier2 image shows a pattern similar to that for my quadratic trend in most (8/11) of the volumes. I assume this means that many of my group 2 subjects are being deweighted as outliers, but what I am curious about is how the same region can have a near identical (if not identical) pattern for MOST of the subjects in this group as outliers.
My second question is if this issue with the prob_outlier image can lead to strange results I mention above regarding the quadratic trend.
Thanks for your help,
david paulsen
|