Dear SPMers,
we would like to set up a flexible factorial design (5x3 ANOVA) in spm8 for our fMRI data but are not sure what is the best way to do it since the study design is quite complex. Especially we’re not sure which comparions are possible and how to define the contrast weights.
We have 5 groups (one is placebo, the other four groups received different types of medication) with 3 time points each (T1 was baseline, T2 after administration of placebo/medication, T3 after additional administration of another substance to all 5 groups). Every group consists of either 6 or 7 subjects.
We would like to test the main effect of group and time and the interaction group x time by also accounting for the factor subject. Using the flexible factorial setup in spm8, we created three factors:
1. subject
2. group (5 levels)
3. time (3 levels)
and then fed 3 scans per subject for T1, T2 and T3 into the model. In the main effects & interactions section, we set up the three main effects, and the interaction between group and time [2 3]. Our question is now how to define the main effect of condition and the main effect of group (variables MEg and MEc like in the Gläscher / Gitelman manual) and how to set up the contrast weights to test for the mentioned main effects and interactions.
We have 4 main questions:
1. How should we define MEc, when we hypothesize T2>T1 and T2>T3 (and no differences between T1 and T3)? Is this contrast correctly defined: [-1 2 -1]? How would MEc look like with the hypothesis T2>T1 an T3>T1? [-2 1 1]?
2. When finding an interaction effect group x time: Is it possible to evaluate the post hoc tests in the same model? (i.e. change over time in single groups; and comparison of this change between groups) Or do we have to set up a separate model for main effect of time in a single group and comparison of groups? Which post hoc test would be appropriate?
3. Does the 5x3 flexible factorial design make sense at all, especially because of the small sample size? Alternatively we could try a 3x3 design (by merging medications with similar mechanisms of action). So after consolidation of groups we would have the following situation: Placebo vs Group A vs Group B.
Does anybody have experience in computing the contrast weights for this case? The hypothesis for the main effect of time is the same as in the 5x3 design above.
4. Would a full factorial be appropriate instead of a flexible factorial with this within-between subject design?
Any help would be greatly appreciated.
Thanks in advance,
Elisabeth
########################################################################
To unsubscribe from the SPM list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=SPM&A=1
|