R Mice: analyzing imputed dataset with with() function

Question

I am trying to run an analysis using a dataset imputed through the mice package. However, when using the with() function to run a linear regression on the imputed data, I get an error stating that the variables I am trying to analyze do not exit. Here is an example:

imputed_dataset <- mice(original_data, formulas = list_of_formulas, m = 5, maxit = 10)
model <- with(imputed_dataset, lm(my_outcome ~ some_imputed_predictor + other_predictor_that_was_not_imputed))

I can get the with() function to return a model if I only include predictors that have been imputed by Mice. I am not able to get the function to work if a variable that was not imputed is included in the formula. What should I do to be able to run my regression model, with some imputed predictors and some non-imputed predictors?

Can you make this post reproducible by adding data? Do `dput(original_data)`. — Ronak Shah, Feb 06 '20 at 05:28
Might be safer to include all the variables needed on the RHS of the lm-formula in the mice imputation formula. — IRTFM, Feb 06 '20 at 09:07

score 1 · Answer 1 · answered Feb 06 '20 at 15:26

The answer to this problem can be found in this thread. The formula needs to be wrapped like this

model <- with(imputed_dataset, formula(format(lm(my_outcome ~ my_regressors))))

Running mice with a formula as a variable: instant evaluation instead of later evaluation?

R Mice: analyzing imputed dataset with with() function

1 Answers1