Randomization Tests in Switchback Experiments

TL;DR

This paper introduces a finite-sample valid, distribution-free randomization-test framework for switchback experiments, addressing challenges like serial dependence, seasonality, and heavy tails, without relying on parametric outcome models.

Contribution

It develops a novel conditional randomization test approach that ensures valid p-values in complex time-dependent settings, with diagnostics and power analysis tools.

Findings

Provides finite-sample valid p-values for null hypotheses

Demonstrates favorable size and power in simulations

Offers diagnostics for carryover and non-anticipation

Abstract

Switchback experiments--alternating treatment and control over time--are widely used when unit-level randomization is infeasible, outcomes are aggregated, or user interference is unavoidable. In practice, experimentation must support fast product cycles, so teams often run studies for limited durations and make decisions with modest samples. At the same time, outcomes in these time-indexed settings exhibit serial dependence, seasonality, and occasional heavy-tailed shocks, and temporal interference (carryover or anticipation) can render standard asymptotics and naive randomization tests unreliable. In this paper, we develop a randomization-test framework that delivers finite-sample valid, distribution-free p-values for several null hypotheses of interest using only the known assignment mechanism, without parametric assumptions on the outcome process. For causal effects of interests, we impose two primitive conditions--non-anticipation and a finite carryover horizon m--and construct conditional randomization tests (CRTs) based on an ex ante pooling of design blocks into "sections," which yields a tractable conditional assignment law and ensures imputability of focal outcomes. We provide diagnostics for learning the carryover window and assessing non-anticipation, and we introduce studentized CRTs for a session-wise weak null that accommodates within-session seasonality with asymptotic validity. Power approximations under distributed-lag effects with AR(1) noise guide design and analysis choices, and simulations demonstrate favorable size and power relative to common alternatives. Our framework extends naturally to other time-indexed designs.