1
$\begingroup$

Auxiliary variables may have themselves missing values. According to the following website, one (can) include(s) auxiliary variables also as variables to be imputed. Is this common practice?

https://stats.idre.ucla.edu/stata/seminars/mi_in_stata_pt1_new/

I developed my imputational model without knowing this. It seemed "not right" for me thats why I used auxiliary variables to explain my missing values but not as variables to be imputed themselves.

However, my knowledge on multiple imputation is very limited.

$\endgroup$
1
$\begingroup$

Yes, if there is missingness in the auxiliary values, you should (and have to) impute them as well. If the same individual had missingness both on some focal variable and on the auxiliary variable, how could the auxiliary variable be used to impute their focal variable value? The auxiliary variable has to be imputed so it can be used to impute other values. With a lot of missingness in the auxiliary variable, it will probably do more harm than good because many of the imputed values of the focal variable will come from high-variability imputations of the auxiliary variable.

$\endgroup$
  • $\begingroup$ Thank you for your reply. I was thinking about excluding cases (listwise deletion) that have missings in my auxiliary variable and conducting the imputation model and substantive model on the remaining cases. Isn't this a more conservative approach? $\endgroup$ – user18075 Jun 14 at 18:50
  • $\begingroup$ Why would it be? Latest deletion requires a very strong MCAR assumption. $\endgroup$ – Björn Jun 14 at 19:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.