Nonparametric Estimation of Heterogeneous Treatment Effects: From Theory to Learning Algorithms
Abstract
The need to evaluate treatment effectiveness is ubiquitous in most of empirical science, and interest in flexibly investigating effect heterogeneity is growing rapidly. To do so, a multitude of modelagnostic, nonparametric metalearners have been proposed in recent years. Such learners decompose the treatment effect estimation problem into separate subproblems, each solvable using standard supervised learning methods. Choosing between different metalearners in a datadriven manner is difficult, as it requires access to counterfactual information. Therefore, with the ultimate goal of building better understanding of the conditions under which some learners can be expected to perform better than others a priori, we theoretically analyze four broad metalearning strategies which rely on plugin estimation and pseudooutcome regression. We highlight how this theoretical reasoning can be used to guide principled algorithm design and translate our analyses into practice by considering a variety of neural network architectures as baselearners for the discussed metalearning strategies. In a simulation study, we showcase the relative strengths of the learners under different datagenerating processes.
 Publication:

arXiv eprints
 Pub Date:
 January 2021
 DOI:
 10.48550/arXiv.2101.10943
 arXiv:
 arXiv:2101.10943
 Bibcode:
 2021arXiv210110943C
 Keywords:

 Statistics  Machine Learning;
 Computer Science  Machine Learning
 EPrint:
 To appear in the Proceedings of the 24th International Conference on Artificial Intelligence and Statistics (AISTATS) 2021