Dear FSL experts:
I have been reading meticulous answers provided about how GLM should be designed, but not entirely confident that my design is correct. Could I ask for some help in looking over this? Sorry in advance for grammar and spelling mistakes, English is my second language.
1. I have four groups of subjects actually, but I put all of them together as one, after controlling their age and gender, I want to know how their Reading Ability (RA) correlates with some DTI parameters(like FA). So here is my design :
EV1: Reading Ability(Demeaned) EV2: AGE(Demeaned) EV3: SEX(Demeaned)
1 0 0
Does this look right?
2. Still, in the design that I mentioned above, if I want to test the main effect of reading ability, can i add "group" as another EV so that i can partial the effect of GROUP? the design should be like this ?
EV1: Reading Ability(Demeaned) EV2: AGE(Demeaned) EV3: SEX(Demeaned) EV4: GROUP(Demeaned)
1 0 0 0
3. Also, if I want to test the interaction effects of RA * GROUP. can I set up the model like this:
EV1: AGE(Demeaned) EV2: GROUP(demeaned) EV3: Reading Ability (Demeaned) EV4: GROUP*Reading Ability (Demeaned)
0 0 0 1
4. Among these EVs, actuallly AGE has a significant correlation with the participants' Reading Ability, then is it still ok to add these covariates like this ? will there be a collinearity problem? I saw one reply concerning this issue, and the suggestion was to test VIF of these covariates?
Any advice would be greatly appreciated! Thank you very much for your time! ^_^
Best regards,
Zoe.G
|