Dear SPM-experts,
I would like to compare 3 different groups (A, B & C), each comprising a
different number of subjects (nA, nB and nC).
Setting up a one-way ANOVA (1 factor with 3 levels) allows for an
overall comparison (f-test) and post-hoc t-tests between groups (t-tests).
However, now I need to compare groups (A & B) vs C and I am not sure
about the correct t-contrast:
1) is it simply [0.5 0.5 -1]?
2) or do groups A and B require a weighting with the corresponding
sample size in each group, ie. [nA/(nA+nB) nB/(nA+nB) -1]?
Both would give me a balanced contrast, but I am not sure whether the
number of subjects per group are already accounted for (-> contrast 1)
or if one needs to consider this in the contrast setting manually (->
contrast 2)?
I already checked the Glascher and Gitelman-tutorial as well as the SPM
archives, however, I could not find an answer to this (rather simple) issue.
Thank you in advance for your help and kind regards,
Andreas.
--
Andreas Hahn, MSc.
Functional Neuroimaging Group - PET & fMRI
Department of Psychiatry and Psychotherapy
Medical University of Vienna, Austria
Phone: +43 1 40400 3825
Email: [log in to unmask]
Web: http://www.meduniwien.ac.at/neuroimaging/
|