Multivariate Regression Trees:
An Alternative Method
Bruce Ratner, Ph.D.
Statistical modeling is frequently used for making predictions. A common task - predicting a single continuous variable - is performed using the ordinary regression model. However, the task of simultaneous predictions of many continuous variables has not been popular because of the complexity of the multivariate (many-variable) regression model (MRM). The simple alternative method for the MRM, the multivariate regression tree (MAID), is itself not popular because there is no software to perform the method. The purpose of this article is to present a new, readily available alternative method for the MRM - the quasi-MAID - that builds multivariate regression trees. An illustration of the new method is presented to hopefully stimulate interest in performing simultaneous prediction, especially via regression trees.
1 800 DM STAT-1, or e-mail at firstname.lastname@example.org.