Causal Inference from Small High-dimensional Datasets
Abstract: Many methods have been proposed to estimate treatment effects with observational data. Often, the choice of the method considers the application's characteristics, such as type of treatment and outcome, confounding effect, and the complexity of the data. These methods implicitly assume that the sample size is large enough to train such models, especially the neural network-based estimators. What if this is not the case? In this work, we propose Causal-Batle, a methodology to estimate treatment effects in small high-dimensional datasets in the presence of another high-dimensional dataset in the same feature space. We adopt an approach that brings transfer learning techniques into causal inference. Our experiments show that such an approach helps to bring stability to neural network-based methods and improve the treatment effect estimates in small high-dimensional datasets. The code for our method and all our experiments is available at \url{github.com/HiddenForAnonymization}.
Submission Length: Regular submission (no more than 12 pages of main content)
Previous TMLR Submission Url: https://openreview.net/forum?id=J3cDdkFfEU&layout=2&sort=date-desc
Changes Since Last Submission: We fixed the anonymization issue on the supplementary material.
Round 1 - based on the reviewers' comments, we made modifications on the following sections:
Reviewer 1: Sections 1 (removed a sentence), 2 (fixed a citation), 3.5 (clarified how the treatment effect is estimated), 4 (extra explanation on the AIPW estimator adopted)
Reviewer 2: Sections 3.4 (extended the discussion on how to obtain the hyper-parameters)
Reviewer 3: Sections 1 (added an extra motivational example), 3.3 (added a paragraph on why proposed representation is beneficial for the goal of causal inference).
Assigned Action Editor: ~Antti_Honkela1
Submission Number: 440
Loading