I am trying to run CART on multiple-imputed datasets (50 data sets obtained using predictive mean matching with the MICE package). Is it possible to use ‘rpart’ on MI data sets and produce one single decision tree which reflects all the variability of the multiple imputed sets of data? For example:
#MI using MICE imp <- mice(missing_data( , ), m = 1, maxit = 5, donors=5, seed=123, print=F)
#Regression Tree – CART Fulltree<-rpart(imp$outcome1 ~ ., data=imp, method = ‘anova’)