Question about modelling interactions (2 groups, 3 nuisance variables (2 categorical + 1 continuous), and the interaction between a continuous variable and groups

1

Hello All,

I have a question about my design matrix for FSL’s randomise and hope to get your feedback on it.

The question here is to test the interaction between cognitive score and group (Group A and Group B).

I would like to control for study site (Site A and B), Sex and Age.

N = 50

3 groups

  • Site (Site A x Site B)
  • Sex (Males and Females)
  • Group (A and B)

Continuous covariates

  • Age (demeaned)
  • Cognitive score (demeaned)

After looking at some examples on the FSL mailing list, I created the following matrix below:

image
image2994×848 207 KB

Is that correct? Do I need really to split the groups by gender and site? I ask that because I have also seen quite a few other examples where nuisance categorical covariates were simply modelled as 1 and 0 in single columns.

Thanks in advance :slight_smile: