INVESTIGATION OF THREE METHODS OF VARIABLE SELECTION: LINEAR MODEL SELECTION, CART AND TREED REGRESSION

Authors

Todd R. Nelson, Brigham Young University
Dr. Scott Grimshaw, Brigham Young University

Keywords

variable selection, linear model selection, cart, treed regression

College

Physical and Mathematical Sciences

Department

Statistics

Abstract

My project for the Research and Creative Activities Scholarship Award looks at the effectiveness and accuracy of a popular new method of variable selection. The method under investigation uses tree-based models to select important variables and then uses classical least squares estimation to fit a regression model with the variables contained in the tree. This research tests the strengths and weaknesses of this approach to modeling as compared with more classical methods.

Recommended Citation

Nelson, Todd R. and Grimshaw, Dr. Scott (2013) "INVESTIGATION OF THREE METHODS OF VARIABLE SELECTION: LINEAR MODEL SELECTION, CART AND TREED REGRESSION," Journal of Undergraduate Research: Vol. 2013: Iss. 1, Article 2809.
Available at: https://scholarsarchive.byu.edu/jur/vol2013/iss1/2809