Accuracy of random-forest-based imputation of missing data in the presence of non-normality, non-linearity, and interaction

There are caveats of course! The performance of random forests decreases substantially when logistic regression relationships with interactions exist within the data. It appears to still outperform predictive mean matching under most conditions. https://

@Benjaming_G You could use imputation methods based on random forests (see R package https://t.co/SVTgMjIx7p) BUT you are going to get biased regression coefficient estimates and confidence interval coverages that are too small (see https://t.co/kuVpzfCfkZ

https://t.co/7bpwuXUGRG