Multiple imputation of categorical/binary variables

[Georgia Petrou]

Hi all,

I am currently trying to run a multiple imputation analysis with all of the variables having missing data (exposure, outcome, modifier and confounding variables). All of the variables are categorical or binary.

I used the following command to carry out a multiple imputation analysis:


mi set mlong

mi register imputed gad7_binary Nfear_co gender_group_codes age_group_codes hs_group_codes prof1_group_codes education_level current_emptstatus PPE_amount

mi impute chained (regress) gender_group_codes age_group_codes hs_group_codes prof1_group_codes (ologit) education_level current_emptstatus (ologit) PPE_amount (mlogit) Nfear_co=brs_binary, add(20)

It did give the following error:

Performing chained iterations ...
hs_group_codes: missing imputed values produced
This may occur when imputation variables are used as independent variables or when independent variables contain missing values. You can specify
option force if you wish to proceed anyway.
r(498);

I used the force option as suggested, and while that allowed the multiple imputation analysis to continue, it did not impute all the missing values but only a few from each variable.

Is there a reason why that happened?

Any suggestions of how I could resolve this issue?


Thank you in advance.