Historical regression trees are an extension of standard trees, producing a non-parametric estimate of how the response depends on all of its prior realizations as well as that of any time-varying predictor variables. The method applies equally to regularly as well as irregularly sampled data. The package implements random forest and boosting ensembles based on historical regression trees, suitable for longitudinal data. Standard error estimation and Z-score variable importance is also implemented.
copied from cf-post-staging / r-htree