Unbiased regression trees for longitudinal and clustered data

We generalize the simulation design in sections 3.2 and 3.3 of the paper in order to test how sensitive the proposed unbiased RE-EM tree method is to unequal sample sizes per object. Recall in sections 3.2 and 3.3 of the paper there are 100 individuals and 5 observations for each individual in the training set and 50 observations for each individual in the testing set. Here, we have four types of unbalanced/balanced data structure,