Hi everyone,
I’m using FSL’s GLM and I’d like to double-check that I’m coding a binary covariate (sex) correctly.
I have two types of analyses (see attached design screenshots):
- Between-group comparison:
- 3 groups (each with its own dummy group EV)
- Covariates: sex (female/male), age, years of education, time since stroke
- Cross-sectional study with a single group:
- Only one group (one EV and all participants set to 1)
- Lifestyle variables of interest: IPAQ, SBQ, PSQI, MEDAS
- Covariates: sex (female/male), age
In both designs, I model sex as one EV coded 0/1 for the two categories and then demeaned across subjects. All other covariates are continuous and also demeaned.
According to the FEAT GLM documentation, binary variables can be entered as a single 0/1 EV, but I wanted to confirm that this is the recommended / correct approach in these multi-covariate, multi-group designs.
My question:
Is it appropriate in FSL to model sex as a single demeaned 0/1 EV (rather than two EVs) in these kinds of GLMs, or would you advise a different coding?
Thank you very much for your help!
Best,
Adrià Bermudo-Gallaguet
PhD Candidate in Clinical and Health Psychology
University of Barcelona