Hi there,
I have 3 variables in my dataset of 970 patients, for which I have missing data. Two of the variables are baseline and follow-up measurements of the same biomarker and the 3rd variable is a biomarker which is highly correlated with the previously referred to biomarker. As these biomarkers are not normally distributed, I think it would be most appropriate to use predictive mean matching to impute values for missing observations. However, my concern is whether it is appropriate to use PMM to impute on all three variables, given that it's a univariate imputation method? Any advice on the above would be greatly appreciated.
Thanks,
Claire
I have 3 variables in my dataset of 970 patients, for which I have missing data. Two of the variables are baseline and follow-up measurements of the same biomarker and the 3rd variable is a biomarker which is highly correlated with the previously referred to biomarker. As these biomarkers are not normally distributed, I think it would be most appropriate to use predictive mean matching to impute values for missing observations. However, my concern is whether it is appropriate to use PMM to impute on all three variables, given that it's a univariate imputation method? Any advice on the above would be greatly appreciated.
Thanks,
Claire
