Applications of Regression Trees in Survey Data Analysis

• Activity Number: 207
• Type: Roundtables
• Date/Time: Monday, August 10, 2015 : 12:30 PM to 1:50 PM
• Sponsor: Government Statistics Section
• Author(s): Daniell Toth*
• Companies: Bureau of Labor Statistics
• Keywords: recursive partitioning ; sample design ; nonparametric regression ; nonresponse

Abstract

Since the idea of regression tree models was introduced in the statistics literature as a way to understand complex interactions among variables, they have grown in popularity due to their easily interpreted structure. This structure lends itself to many applications in survey data analysis. Survey data is usually collected through a complex sample design, so methods are needed to incorporate this design information in the tree construction. This roundtable discussion will focus on methods for constructing regression trees and applications in survey research. We also will consider software that provides recursive partitioning algorithms, including packages in R and programs available from the leader of this discussion.