Dear Guillaume and members of the list,
in follow-up to my previous post
https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind1610&L=spm&P=R40094&1=spm&9=A&J=on&d=No+Match%3BMatch%3BMatches&z=4
I would like to ask you which is the appropriate approach to a posthoc t-test in the block-diagonal design that you suggested me in order to make spm process partitioned error terms. You specifically suggested me to create for each subject either a contrast 'average effect of condition' or difference between conditions and then bring the cons to a one-way anova or a big one sample t-test. So I was wondering whether it would be most appropriate to make the t-tests in the one-way anova, and so mostly out of 'average effect of condition' con images, or whether I should create another individual con image, for instance, an unidirectional t-test, and create another, independent t-test, at the 2nd level?
Also, I was wondering if I can or shall make the block-diagonal design acknowledge the inequalities in my samples size (37, 40, 60)?
Looking forward your feedback,
Bien cordialement,
Isadora
|