A Comparison of Machine Learning Methods to Estimate Conditional Treatment Effect With Multilevel Observational Data

• In Event: Friday Roundtable Session 4:55 pm
  In Roundtable Session: Machine Learning, Statistical Software, and Data-Driven Analytics (Table 32)

Abstract

The objective of this project is to further assess the potential of the popular machine learning methods in dealing with multilevel observational data for their ability to estimate the conditional treatment effect. The two methods of interests are the causal forest and the Bayesian additive regression trees. These two tree-based machines learning gained wide applications, but were not comprehensively compared in the multilevel conditions where the treatment was assigned at the individual level, in addition with extra distractors. Through manipulating simulation conditions, the project wish to show applied researchers the performed of each method in each condition when incorporating propensity scores. The significance of study and future studies’ direction are also discussed.

Authors

• Jia Quan, University of Kansas

• Walter L. Leite, University of Florida

• Wei Li, University of Florida