Non-shared instruments should be able to model some pleiotropy
Simplest approach - zero-relevance instruments associating with the outcome more than expected by chance: measure of global pleiotropy; conclusion that using multiple ancestries allows us to adjust for pleiotropy using SEM whilst determining how unreliable current instruments might be in single ancestry analysis
More complex - Try to adjust for estimated pleiotropy
Complexity is that SEM seems to handle directional pleiotropy, so is this even required?
Complexity is that SEM seems to handle directional pleiotropy, so is this even required?