Abstract

GUIDE is a classification and regression tree algorithm that can be used in many missing data settings. For prediction and function estimation, GUIDE can yield predicted values without imputation of missing values in predictor variables. For the purpose of missing value imputation itself (e.g., as a data completion step prior to application of methods that require completely observed data), GUIDE does not require imputation initialization nor iteration, unlike sequential imputation methods such as MICE and missForest. The talk will present the key ideas and some comparative results on real and simulated data.