Regression trees for longitudinal and multiresponse data
This work addresses limitations in regression tree modeling for longitudinal and multiresponse data, offering an unbiased and flexible method for researchers in statistics and data analysis, though it is incremental as it builds on existing GUIDE concepts.
The authors tackled the problem of selection biases and computational difficulties in regression tree models for longitudinal and multiresponse data by proposing an alternative based on the GUIDE approach, which treats data series as curves and uses chi-squared tests for splitting, resulting in unbiasedness and applicability to various data conditions, with simulation results showing mean squared prediction error comparisons.
Previous algorithms for constructing regression tree models for longitudinal and multiresponse data have mostly followed the CART approach. Consequently, they inherit the same selection biases and computational difficulties as CART. We propose an alternative, based on the GUIDE approach, that treats each longitudinal data series as a curve and uses chi-squared tests of the residual curve patterns to select a variable to split each node of the tree. Besides being unbiased, the method is applicable to data with fixed and random time points and with missing values in the response or predictor variables. Simulation results comparing its mean squared prediction error with that of MVPART are given, as well as examples comparing it with standard linear mixed effects and generalized estimating equation models. Conditions for asymptotic consistency of regression tree function estimates are also given.