I have been thinking about scenarios under which it makes sense to use imputation for prediction models and am struggling to come up with a case. Yikes! Even for inference, as long as you do some doubly robust approach, I’m not sure I see the value (other than for precision, but then is no longer a question of bias and thus is a question for a different day!)
- When is complete case analysis unbiased?
I have been thinking about scenarios under which it makes sense to use imputation for prediction models and am struggling to come up with a case. Yikes! Even for inference, as long as you do some doubly robust approach, I’m not sure I see the value (other than for precision, but then is no longer a question of bias and thus is a question for a different day!)
https://livefreeordichotomize.com/posts/2023-04-28-imputation-might-be-silly/