Dear FSL experts,
I only just realised that for some reason the message didn't show, sorry! Hopefully this time it will work.
I am very puzzled because I ran two analyses that (from my understanding) should yield equivalent results but in fact they don't. I did seed-based resting state analyses and then wanted to test whether functional connectivity of my seed (fusiform gyrus) with other brain areas is associated with performance in a cognitive test. I set up a onesample t-test with covariate GLM (see attachment of my first empty message), then ran FLAMEO / easythresh to test for a positive and a negative relationship between BOLD activation (function connectivity) and a behavioral covariate (cognitive test).
To varify, I used the resulting cluster mask (of the are where I had found sig. results) and extracted parameter estimates from all individual participants from that area (one coherent cluster). However, when I ran a pearson correlation in SPSS to look at the association between those parameter estimates and the cognitive test results, there is absolutely no significant association (r=.009, p=.9).
I checked my analyses for silly mistakes in the process, but could not find any so I guess there must be a more fundamental error in my approach. Did I do something wrong when setting up the GLM or is there another reason why the results of both approaches could be so different?
Thank you very much in advance for your help!!
Susanne
|