Dear Kelvin > We have 30 subjects and 30 controls who are tested with the same fMRI > paradigm. Individual results are analysed (60 con*.img). To analyse > the group effect, we performed a two sample t test (subject group and > control group are different) by selecting first 30 subjects from test > group (con*.img) and selecting the next 30 subjects from control > group. > > In the group definition, we entered 30 '1' and 30 '2'. After all the > selections (all none/omit), we entered a contrast 1 -1 to evaluate the > group effect. Is that true that the contrast is a population inference > (RFX) rather than a FFX within the 60 subjects we collected? We'd like > to do a RFX. This corresponds to a RFX analysis because the only source of error is provided by the between subject variability in contrasts. > Another question is: If we have a single batch of subjects who undergo > a treatment and fMRI were acquired before and after the treatment. > What kind of test should I do to see if the treatment is effective to > the group in a whole and to the population (if possible)? A FFX analysis would tell you whether the treament was effective in that group and an analysis similar to that above (but using a paired t test at the second level) would emulate a RFX analysis and allow generalization to the population from which the subjects were sampled. I hope this helps - Karl