What are the chances that HAIL at some point will include sandwich corrected SE’s for robust inference in dataset which contain family members?
Best,
Michel
Hi Michel,
This is something which I don’t expect to ever be super high priority for the development team, but that doesn’t mean Hail will never include a sandwich correction – in an open-source project, features arrive when someone adds them!
We do have a linear mixed regression method, which should account for family structure in a first-class way. Is there a particular reason that this model won’t work for your analysis?
Thought I would bring this back up as I have run into another use case for robust/Huber-White/sandwich standard errors recently. Specifically, they seem to be very helpful in combatting heteroscedasticity in the exposure/outcome relationship for GxE interaction tests (for example, see here).
Thanks!
Kenny