Designing Transportable Experiments

TL;DR

This paper introduces a new method for designing randomized experiments that explicitly accounts for the target population, improving the accuracy of treatment effect estimates under covariate shift.

Contribution

It proposes a novel experiment design approach called Target Balance that reduces variance by considering the target population during the design phase.

Findings

Target Balance achieves higher variance reduction asymptotically.

The method reduces variance even with small sample sizes.

The approach provides unbiased ATE estimates under covariate shift.

Abstract

We consider the problem of designing a randomized experiment on a source population to estimate the Average Treatment Effect (ATE) on a target population. We propose a novel approach which explicitly considers the target when designing the experiment on the source. Under the covariate shift assumption, we design an unbiased importance-weighted estimator for the target population's ATE. To reduce the variance of our estimator, we design a covariate balance condition (Target Balance) between the treatment and control groups based on the target population. We show that Target Balance achieves a higher variance reduction asymptotically than methods that do not consider the target population during the design phase. Our experiments illustrate that Target Balance reduces the variance even for small sample sizes.