is there a test to see if the InSIDE assumption holds? If the assumption is that the SNP-exposure effects aren’t correlated with the SNP-outcome effects can’t we just test that empirically?
Replying to “is there a test to see if the InSIDE assumption ho…”:
Unfortunately only in theory without making some strong assumptions. The SNP-outcome associations are a mixture of the direct and mediated effects and it’s difficult to pull those two things apart, so your SNP-outcome estimate doesn’t just represent direct effects.