On Fri, Jan 31, 2014 at 2:36 AM, David Pedrosa <[log in to unmask]>wrote: > Dear all, > > I have a question regarding the analyses of PET data with SPM. We have 25 > subjects in three groups and they were measured during 3 distinct > conditions. Besides, these conditions consisted of up to three repetitions. > Once you average the three repeats, you can use the resulting average images as the inputs into the statistical model. > Now we would like to test for significant differences in the activation > between the three groups but also for differences within the groups with > SPM. > Since you have a repeated measures ANOVA, you can only look at group*condition interactions, you can't look at if the groups are different for each condition. For that you would need a one-way ANOVA for each condition. You can look for condition differences within each group. > So we have a 3x3 repeated measures ANOVA. As I am not very experienced > with GLM in SPM I have been searching this forum. Yet, i have some > remaining questions. First, what are the advantages of using GLMFlex? There > have been several threads about this but I didn't really understood the > point about it. > GLM Flex would give you the group effects by condition. > And second, using a flexible factorial design (if possible), is there any > tutorial or explanation for the contrasts to be used? > If you search S1G1C1 on the list, you will find an example of how to build contrasts. This will work for any design. > Applying the one of gläscher et al. was helpful. For this we defined three > factors (subject, group, condition). > The main effect of group, group comparisons, and individual conditions tested against 0 are invalid statistical contrasts. The old document hasn't been updated to reflect this issue. > However,we are not sure how to define contrasts to detect interactions. We > are looking for a help to understand what we have to do to answer our > hypotheses and would be grateful for any kind of help like links to e.g. > comprehensive tutorials or something. > See the examples in the list by searching for S1G1C1. > Thanks in advance > > David >