Chained Imputation - Auxiliary variables

[Hady Sayed]

Hello

I'm trying to do MICE imputation in Stata.

I have 17 variables in the analysis model and I have to include them all in the imputation model. This means that I would have to find multiple auxiliary variables since no one variable will correlate with all 17.

Is it okay to run the model without having auxiliary variables? I noticed that in the UCLA Stata multiple imputation manual, it says to identify auxiliary variables "if necessary"

Thanks

[Felix Bittmann]

Sure. Without auxil vars, MICE will impute all vars in the model mutually. Auxil vars are helpful if they are strong predictors of your variables that have missing values but you do not want to use them in your analytical models. If there are none available just impute the vars you have,