 |
 |
Dear Alessandro and Francesco, > Sorry to bother you on your personal e-mail address. We are two > young-fellows working at the NIH and struggling with covariates > analyses in SPM99. > We were wondering if you could give us a suggestion about our study. > > We are conducting an aging study with two groups (young and old people) > performing the same task (rest-motor). We already made a first level > analysis and now we have a single con*.img for each subject. Since we > were interested in looking at the effect of a covariate (reaction time) > on our data we moved to a second level analysis using Simple Regression > model (under Basic Models) with raw reaction time values as the > covariate for each subject. Is this correct? Yes it is. The RTs will be automatically/implicitly mean corrected and you will get a random effects inference about the main effect of RT on activation (assuming you have taken the mean RT for each subject as the covariate). > We want also to look at the difference between the two groups. Could > you please suggest us the best way to do this? This is simply done by modeling the RT separately for each group (young and old). You could do this by hand (by mean correcting each groups RTs and entering 2 regressors, one for each group). This is done automatically if you pretend each group is a subject in PET designs and ask for subject x covariate interactions. The interaction is the difference between the two groups in terms of the regression of rCBF on RT and obtains by using the contrast [-1 1]. Note the first analysis is the same as using a contrast [1 1] in the second analysis. I hope this helps - Karl