Hi Vasilis,
Firstly, I would direct you to my recent paper to read about how to correctly implemented these kinds of complex repeated measures models in the GLM (including covariates). There is an example in there of a within x within x between model in SPM which should help you to make sure you are doing things correctly: https://www.frontiersin.org/articles/10.3389/fnins.2019.00352/full
Secondly, I would strongly caution against your approach with covariates. The traditional ANCOVA approach is based on have a covariate that is related to the outcome variable but NOT related to the predictor variables, whereas your approach is engineering covariates that are guaranteed to be related to the groups. The issue is that you can create spurious group differences by doing this (known as Lord's paradox). Miller & Chapman (2001) is the go-to reference for this: https://psycnet.apa.org/record/2001-17627-005.
Please read this carefully and then consider your strategy.
Best wishes,
Martyn
|