Dear Doreen,
This is fine, as your model is that there is a group mean + some effect of age + some effect of depression (but zero effect in the control group for depression). So this model manages to do all of that, and the demeaning is correct too (in order to separate things as cleanly as possible).
All the best,
Mark
On 27 Mar 2013, at 13:03, Doreen O <[log in to unmask]> wrote:
> Hello,
> I am trying to use TBSS to compare FA maps in 3 groups: healthy controls, treatment remitters, and treatment non-remitters. I have a few covariates to include, but it gets complicated because one covariate is age (data available for all subjects), and another is depression severity (data available for 2 groups only). Here is how I have set up my matrix so far:
>
> controls: remitters: nonremitters: age (demeaned) dep severity (demeaned)
> 1 0 0 7.86 0
> 1 0 0 -5.62 0
> 0 1 0 1.20 1.16
> 0 1 0 12.10 -5.84
> 0 0 1 -1.54 7.25
> 0 0 1 2.20 -2.11
>
> The age column is demeaned using the mean of all three groups. The dep severity column is demeaned using the mean of the two patient groups (remitters and nonremitters) and I put zeros in for control values (since they do not have this data). Is this correct?
> Best,
> Doreen O.
|