Hi everybody,
we are new to GLM and have a question to a GLM-Design. Could you maybe help us?
We want to control our DTI-data for gender.
male subjects are marked with 0
female subjects are marked with 1
Our design matrix would look like this:
EV 1: controls
EV 2: patients
EV 3 is gender demeaned of all subjects
Every subjects is in one group
Group EV1 EV2 EV3
1 1 0 gender demeaned
....
1 0 1 gender demeaned
In this case: is demeaning right? Or should we only take the variables 0 and 1?
The contrast would be:
con1 group diff (controls-patients)
con2 group diff (patients-controls)
1 -1 0
-1 1 0
Would you agree with this contrast? Or would you choose an other contrast with 4 EVs ( geneder for each group in a different EV-column).
If we would do the same for education-years: would you demean education-years in this case?
Hoping for some answers and gratefully,
Martha
|