A conditional randomization test to account for covariate imbalance in\n randomized experiments

2015/10/22 by Jonathan Hennessy, Hennessy, Jonathan, Tirthankar Dasgupta +7 · 1 citation
Agricultural and Biological Sciences · Mathematics · #FOS: Computer and information sciences #Methodology (stat.ME) #Pesticide Residue Analysis and Safety #Statistical Methods in Clinical Trials

paper · pdf · doi:10.48550/arxiv.1510.06817

openalex publication_date 2015/10/22 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

We consider the conditional randomization test as a way to account for\ncovariate imbalance in randomized experiments. The test accounts for covariate\nimbalance by comparing the observed test statistic to the null distribution of\nthe test statistic conditional on the observed covariate imbalance. We prove\nthat the conditional randomization test has the correct significance level and\nintroduce original notation to describe covariate balance more formally.\nThrough simulation, we verify that conditional randomization tests behave like\nmore traditional forms of covariate adjustmet but have the added benefit of\nhaving the correct conditional significance level. Finally, we apply the\napproach to a randomized product marketing experiment where covariate\ninformation was collected after randomization.\n

Cited by

• Remote Auditing: Design-based Tests of Randomization, Selection, and Missingness with Broadly Accessible Satellite Imagery

Related

• A Randomization-Based Theory for Preliminary Testing of Covariate Balance in Controlled Trials
• Conditional Randomization Rank Test
• Covariate Shift Corrected Conditional Randomization Test
• Inference in experiments conditional on observed imbalances in covariates
• Revisiting Randomization with the Cube Method