Unbiased Regression Trees for Longitudinal and Clustered Data

This page contains material related to the paper "Unbiased Regression Trees for Longitudinal and Clustered Data," by Wei Fu and Jeffrey S. Simonoff, which appeared in 2015 in Computational Statistics and Data Analysis.

Click here to access an earlier version of the paper on the Social Science Research Network (SSRN) site.

Click here for supplemental material giving more details on some of the simulations done in the paper.

R code is available to construct an unbiased RE-EM tree. First, the R package that fits the CART-based RE-EM tree should be installed and loaded; it is available at CRAN. The current version of the package is 0.90.4, and any earlier versions should be updated or replaced. The party package should also be installed and loaded. Then, this file should be downloaded to your computer, and the REEMctree function should be copy and pasted into R. After saving the workspace this function will then be available to construct the unbiased RE-EM tree. The Tree attribute of the object created is a party object, and can be plotted using that package's associated plot command, predictions can be made using the packages predict command, and so on.