Print

Print
Hi Sean,

Neither of these will test the interaction I'm afraid. You'd have to set up a model with a 3rd column that is the product of EV1 and EV2. Also, add an intercept (EV4). The contrasts to test the intercept would then be [0 0 1 0]' and [0 0 -1 0]'.

Mean-centering EV1 and EV2 before computing EV3 is in general a good idea. Although it won't affect the results for the interaction, it will affect the coefficients for the main effects, even with the intercept modelled, and in general it makes these more easily interpretable.

All the best,

Anderson



On 9 May 2014 05:08, Sean Hatton <[log in to unmask]> wrote:
Hi all,

This is a simple question but unsure of the interpretation. I have one group in TBSS and want to look at the interaction of two covariates on FA. So I have

EV1:Cov1     EV2:Cov 2
74                    25
42                   24
55                   22
etc                  etc

From what I have gather from the GLM setup instructions I should have my contrasts as:

Title   EV1    EV2
C1    +1       -1
C2    -1        +1

Unsure what the interpretation of this would be. Would a significant C1 cluster represent covariate 1 has a positive correlation with FA? Or that C2 has a negative correlation with FA? Should I set it up as:

Title   EV1    EV2
C1    +1       +1      <--- The interaction of covariates is positively associated with FA
C2    -1        -1      <--- The interaction of covariates is negatively associated with FA

Cheers,

Sean Hatton
Brain & Mind Research Institute
University of Sydney