hi spm
a question about how to use the PET/SPECT models>conditions and covariates
option to compare groups of patients for a VBM study.
I have 3 patient groups, and variables which are sometimes nuisance (eg
age) and sometimes interesting (eg memory). I want to know: of the voxels
which are associated with an interesting variable eg memory score, which
have greater height values for groups A and B vs group C (ie C is the
disease group)?
when I enter all conditions and covariates and ask for an interaction, I
get 7 columns in my design matrix - the 3 groups, 3 columns which look
like covariate columns and the intercept (+ nuisance variables but let's
not complicate the issue). specific question:
which weights are needed?
the following seems likely where the groups are ordered A B C
0 0 0 1 1 -2
or perhaps
1 1 -2 1 1 -2
are either of these right?
chris summerfield
dept. psychology
columbia university
|