453 – Recursive Partitioning for Modeling Survey Data and Randomized Trials
A Comparison of Classification and Regression Tree Methodologies When Modeling Survey Nonresponse
William Cecere
Westat
When computing survey weights for use during the analysis of complex sample survey data, an adjustment for nonresponse is often performed to reduce the bias of the estimates. Many algorithms and methodologies are available to analysts for modeling survey nonresponse for these adjustments. Lohr et al. (2015) discussed the benefits of using classification trees for estimating response propensities in surveys and how these methods could be used to reduce nonresponse bias. In this paper, we extend their findings and recommendations based on expanded simulations for more complex sample designs, such as a stratified design with equal sample size allocation. We evaluate the effect of some classification tree-based methods on the reduction of nonresponse bias and investigate the performance of the methods when they are used to adjust survey weights. We discuss the benefits of using these methods for estimating response propensities in surveys.