Dear statistician
The result of a T-test between two conditions on a secondlevel analysis differs from the result of exact the same contrast of a post-hoc T-test in a factorial design (ANOVA). Why? (In many cases the post-hoc tests provide stronger effects)
I'm only interested in T-tests but because I screen a lot of contrasts I collect all con images in a factorial design to compare all contrasts using the same SPM design. I did not expect that such post-hoc tests are different than using standard t-Test design.
However, I do notice is that the degree of freedom as labeled text to maximum intensity projections ( e.g. SPM{T102} ) are different, and much higher in the post-hoc tests.
Thanks for any comments
RL
|