85fc37b18c57097425b52fc7afbb6969-AuthorFeedback.pdf

We thank each reviewer for their insightful and constructive feedback. It will surely improve the manuscript.1 Reviewers #2 and #4 suggested that we illustrate our the-2 ory with experiments. In Figure 1, we show the outcome of one such experiment. Indeed,18 CART trees are invariant to strictly monotone transformations of the in-19 dividual predictor variables. Reviewer #4 asked for clarification on whether the44 quantity in Assumption 1 governs the convergence rate of both the training error and prediction error. Assumption45 1 is merely a technical condition about the maximum number of data points in each node and enables one to use46 the correlation comparison inequality inFact 2--it does not explicitly control the training and prediction errors.