We had something come up in some of our data that I was hoping for some explanation/clarification of from the list.
We wanted to run a 1-sample T-test with a covariate to control for average Fusiform Face Area activation. We first ran the model with no covariates, and extracted a 5mm-radius VOI from the peak Fusiform activation.
We then ran a 1-sample T-test, adding a covariate with the extracted Fusiform values. The covariate was overall mean centered (the default option).
Choosing a contrast of [1 0] in the resulting model (1 over the constant column, 0 over the covariate) should -- according to a previous post by Tom Nichols here: https://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=ind0803&L=SPM&D=0&I=-3&d=No+Match%3BMatch%3BMatches&m=30570&P=81801 give us the "expected response for an individual with average fusiform while accounting for linear effect of fusiform" since our covariate is centered.
In the resulting T-map (dof=21), there is a very strong peak exactly at our Fusiform VOI (the rest of the activations range from about T=3.5 to T=8, while the Fusiform is T = 34). At first I thought I'd just mixed up my contrast and was just looking at the [0 1] contrast, but I've triple checked everything. The [0 1] contrast, as expected, also shows a very strong effect in the Fusiform VOI (T = 18, compared to other brain regions ranging from T = 3.5 to about 5).
Can someone help us to understand what might be going on here? We're just confused as to why attempting to control for activation in a particular brain region would result in a very large effect in that exact region.
-Mike Angstadt
**********************************************************
Electronic Mail is not secure, may not be read every day, and should not be used for urgent or sensitive issues
|