Hi Walid,

Please, see below:

On 31 July 2016 at 06:10, Walid Y. <[log in to unmask]> wrote:

Hello FSL experts,

I am trying to create a design where i want to check the difference between 2 groups patient and control (TBSS/Glm).

However i want 1 EV to be a covariate of no interest and two of interest.

Currently I’m at:

EVs:

G1 G2 Age Variable1 Variable2 

1 0 demeand demeand demeand

0 1 demeand demeand demeand

Contrasts: 

C1 1 -1 0 0 0

C2 -1 1 0 0 0

C3 0 0 0 1 0 

C4 0 0 0 0 1

C5 0 0 0 0.5 0.5

Contrasts C1-C4 are fine, and the interpretation as indicated below is correct. C5 can only be used if the measurement units of variables 1 and 2 are the same and their sum is meaningful. I am not sure what is your hypothesis for C5 but perhaps an F-test with C3 and C4 (for any effect) or a conjunction (for an effect in both) could be better/more informative.

 

I want to test the difference between groups 1 and group 2 taking age as a nuisance covariate and variable 1 and variable 2 covariates of interest.

C3 is the effect of V1 between groups 1 & 2 (taking age and V2 as nuisance covariates)

C4 is the effect of V2 between groups 1 & 2 (taking age and V1 as nuisance covariates)

C5 is the effect of both V1 and V2 between groups 1 & 2 taking age as nuisance covariate.

Incase i want to check the effect of V1 with age as a nuisance covariance ONLY, then i will remove V2 from EVs and run it as:

G1 G2 Age Variable1 

1 0 demeand demeand

0 1 demeand demeand

Contrasts: 

C1  1 -1 0 0 

C2  -1 1 0 0 

C3 0 0 0 1  <= V1 with age as a nuisance covariance ONLY

This is fine.

All the best,

Anderson

 

Is my design correct? Thank you for your advice.

Best regards,

Walid