2 Comments
User's avatar
Dean Eckles's avatar

Maybe this is all consistent with the recommendation to not use odds ratios as estimands? Or is it just that it is impossible to convince researchers to stop using them?

Expand full comment
Tim Morris's avatar

This affects collapsible measures too! It’s just that there’s a safety net: if we think we’re targeting a conditional measure and forcing it to be the same for all x then at least readers can be reassured that it’s targeting the marginal measure.

Expand full comment