Dear Martin,
my problem is similar to Eun-Jung's: I have a one-way between-subject ANOVA: 3 groups, 1 dependent variable, but I also have a covariate. The covariate is the demeaned gray matter volume of each subject, which I want to treat as an implicit regressor (thus always giving it a weighting of 0).
My design matrix has 4 columns, columns 1-3 are the groups, column 4 is the covariate. How do I specify my contrasts and especially, what are the appropriate contrast weights? (I have read that positive weights should sum to 1, while group comparisons should sum to 0).
E.g. if I wanted to look at the main effect of my task (across groups), can I use a t-contrast vector like this: [0.333 0.333 0.333 0], or is it [1 1 1 0] ?
Similarly, when looking at the main effect of my task on the first two groups, is it [0.5 0.5 0 0], or [1 1 0 0]?
Presumably, comparing group 1 against group 3 would then be [1 0 -1 0] ?
I'm very grateful for your advice.
Thank you and Kind Regards,
Gaby
|