Dear List,
I have 2 covariates A and B for multiple regression in a 2nd-level
analysis. If I have contrast [1 0] for [A B], I'm basically looking
for correlation with covariate A, while treating B as a nuisance
variable, am I correct? Also, does contrast [1 1] show activation
that correlates with both A AND B, or activation that correlates with
A and/or B and/or (A AND B)? What would be a proper contrast to use
if one is to look for activation that correlates with both A AND B?
A related question is, if there is a negative linear correlation
between A and B, then one would expect to see less activation in
the regression using contrast [1 1] than in the regression using
contrast [1 0] or [0 1], right? This does not appear to be what I
have observed.
Tahnks in advance for any help.
C.-S. Ray Li
Yale Psychiatry
|