Hi there,

I have a data set which includes:
5 polygenic scores for different mental illnesses and age and sex as predictors
16 dependent variables

I have analyzed the relationships between polygenic scores and dependent variables in univariate and "full" (all predictors included) models using mixed effects linear regression. I understand that for so many tests multiple comparisons need to be controlled for. I did so using the FDR method for the p values from the univariate models.

However, for the full models, it was suggested to me that using FDR on multiple values from one model may not be appropriate, and that an omnibus test of significance of adding all polygenic scores using lrtest may be appropriate.

For some of my dependent variables, there was a strong effect of a polygenic score in the full model, but not in the univariate. I want to explore why this is and whether it could be a false positive.

For each dependent variable, I have gone ahead and compared the model including only age and sex to the model that adds all 5 polygenic scores at the same time. Some of them yielded a p value < 0.05. What does this mean in the context of significant effects of my predictors on my dependent variables?

I'm a little lost as to how lrtest will help me answer my question of whether the relationships I saw in the full regression models were true positives. Any help with navigating this and interpreting the results of the LR test in this context would be much appreciated. Thank you in advance.