Adjusting for Covariates in the mint.spls

Dear all,

I am using mint.spls for conducting multivariate analyses in a multisite dataset (20 sites) between a set of imaging and non-imaging variables; I have a question about adjusting for covariates while using mint. When would be the most appropriate to regress out the covariates? within each site or on the whole dataset? Appreciate any leads on this

Best,
Amir

Dear Amir,
We use MINT in a specific case where we need to ‘collate’ different data sets or studies that are independent. MINT is design to then take into account the study effect. In your case the method will account for your sites, across all data sets.

You can also do additional analyses per site using a spls method. The question would be different though. Note that none of our methods can currently handle covariates per sé.

Kim-Anh

Hi and Thank you for the prompt reply. I do understand that mint per se cannot handle covariates. What I meant was how do you suggest treating covariates Prior to the mint analysis? Would you regress them out per site or on the whole dataset?

Ah yes, you can treat covariates prior. In MINT the data are centered per site before the analysis starts, so it might be better to do a regression per site. However it might be worthwhile trying both, just to see if my intuition was correct :slight_smile:

Kim-Anh