I am interested in looking at associations between a PRS and multiple phenotypes gathered within the same cohort (cognition, imaging, mental health symptoms).

However, not everyone in the cohort had all phenotypes assessed (there is a ten fold difference in n).

Does it make more sense to create a PRS for the whole eligible cohort, or create separate PRS for each subsample which has information on a given phenotype? e.g. separate PRS for those with cognition data and for those with imaging data

There could be some difference in who got which sequencing array for example, which might slightly influence the results. But overall I think any differences in PRS would probably be small, and multiple PRS a lot more effort to generate?

Thanks for any advice