Dear colleagues,

I perform a multivariate genetic association study of 10 loci with 3 phenotypes (disorders that can be co-morbid, but they do not necessarily have a causative effect on each other) and other variables like gender. Suppose I want to find effects of the loci on each of the phenotypes, adjusting to others and gender. How should I do that, considering that the dataframe has a notable amount of missing data in genotypes and phenotypes.

The most obvious way, as it seemed to me, was to built models like Ph1 ~ Gene1 + Gender + Ph2 + Ph3... It actually gave results, but are they robust? Gender has the greatest effect due to the great F:M ratio skew. Perhaps due to low study power, CIs for ORs are very wide.

I got a recommendation to regress out gender, and I would also do this with other phenotypes. However, I am not sure, should I take just residuals from m1