"Estimating Treatment Quantiles Without Assumptions"" - Siddharth Bhandari, Research at TTIC

Originally presented on: Friday, April 26, 2024 at 12:30pm CT, TTIC, 6045 S. Kenwood Avenue, 5th Floor, Room 530
Speaker: Siddharth Bhandari, TTIC
Title: "Estimating Treatment Quantiles Without Assumptions"
Abstract: Average Treatment Effect (ATE) estimation is a well-studied problem in causal inference. However, it does not necessarily capture the heterogeneity in the data, and several approaches have been proposed to tackle the issue, including estimating the Quantile Treatment Effects. In the finite population setting containing n individuals, with treatment and control values denoted by the potential outcome vectors a,b, much of the prior work has focused on estimating median(a)− median(b). It is known that estimating the difference of medians is easier than the desired estimand of median(a−b), called the Median Treatment Effect (MTE). The fundamental problem of causal inference -- for every individual i, we can only observe one of the potential outcome values, i.e., either the value ai or bi, but not both, makes estimating the MTE particularly challenging.
In this talk, we will see that the MTE is not estimable, and we will instead focus on a notion of quantile approximation of the MTE. We will prove lower bounds on the approximation factor and see an instance optimal efficient algorithm for approximating the MTE.
Based on joint work with Raghavendra Addanki at Adobe Research (2403.10618.pdf (arxiv.org)).
Tags: #machinelearning #ai #artificialintelligence #computerscience #robotics