Hi all,
I've been playing with ANOVAs, and currently have one set up in which there
is a two-level within subject factor ("region") and a two level across
subjects factor ("group") and subjects themselves are modeled as factors
(see the attached design matrix...ignore the funniness of the matrix at the
bottom...that was the addition of 3 new subjects to group 1). In total there
are 17 in group 1 and 31 in group 2.
What I am interested in doing is to do a valid conjunction analysis (using
Tom Nichols' method) in SPM5. To do so, I have to be able to specify two
valid contrasts, select them both, and this will bring up the conjunction
option. I'd like to evaluate the conjunction of regions (R1-R2) for group 1
with the same contrast for group 2 (rather than doing an overall effect of
region analysis within the ANOVA, since overlap is important to assess and
the sample sizes are quite different).
The problem I'm having, however, is that I can't create valid contrasts for
R1-R2 separately for each region. I've been going off of the contrast
specification article that Jan Glascher and Darren Gittleman wrote,
according to which a contrast for a single regressor of G1R1 for example
might be:
ones(1,14)/14 zeros(1,31) ones(1,3)/14 1 0 1 0 1 0 0 0
By that logic, a contrast between G1R1-G1R2 would be:
ones(1,14)/14 zeros(1,31) ones(1,3)/14 1 0 1 -1 1 -1 0 0
Neither of these gives valid contrasts, in part because neither sum to 0.
I'm not sure how to get around this problem and would appreciate any advice!
thanks,
Amit
|