Calculate averages of imputed data in mice-CodePudding

I would like to calculate the individual-level averages of the imputed datasets in Mice (for presentational purposes, not for analysis). For example, if I have the following dataset with missing.

Id	V1	V2	V3
1	10	NA	9
2	12	8	NA
3	NA	7	11

And 2 imputed datasets like these (imputed values marked with *)

Id	V1	V2	V3
1	10	10*	9
2	12	8	12*
3	11*	7	11

Id	V1	V2	V3
1	10	9*	9
2	12	8	10*
3	14*	7	11

I would like the resulting dataset look like this, with * numbers being the averages of the imputed missing data.

Id	V1	V2	V3
1	10	9.5*	9
2	12	8	11*
3	12.5*	7	11

I can't figure out how to do this in an effective way in R.

CodePudding user response：

You can dirrectly replace the NA values:

 df[is.na(df)] <- ((impute_df1   imput_df2)/2)[is.na(df)]

CodePudding user response：

I brute-forced it.

averages <- complete(imp, 1)
for(m in 2:imp$m){
  averages <- averages   complete(imp, m)
}
averages <- averages / imp$m