Inspection-Guided Randomization: A Flexible and Transparent Restricted Randomization Framework for Better Experimental Design

TL;DR

This paper introduces Inspection-Guided Randomization (IGR), a flexible and transparent framework for restricted randomization that improves causal effect estimation by filtering undesirable assignments based on domain-specific criteria, enhancing reproducibility and robustness.

Contribution

The paper presents IGR, a novel, transparent, and flexible restricted randomization method applicable beyond covariate balance, with pre-registered assignments to prevent p-hacking and improve experimental validity.

Findings

IGR improves effect estimates in simulated education and behavioral health experiments.

IGR effectively filters undesirable treatment assignments based on domain-informed criteria.

Simulation results show IGR outperforms benchmark designs in various experimental settings.

Abstract

Randomized experiments are considered the gold standard for estimating causal effects. However, out of the set of possible randomized assignments, some may be likely to produce poor effect estimates and misleading conclusions. Restricted randomization is an experimental design strategy that filters out undesirable treatment assignments, but its application has primarily been limited to ensuring covariate balance in two-arm studies where the target estimand is the average treatment effect. Other experimental settings with different design desiderata and target effect estimands could also stand to benefit from a restricted randomization approach. We introduce Inspection-Guided Randomization (IGR), a transparent and flexible framework for restricted randomization that filters out undesirable treatment assignments by inspecting assignments against analyst-specified, domain-informed design desiderata. In IGR, the acceptable treatment assignments are locked in ex ante and pre-registered in the trial protocol, thus safeguarding against $p$-hacking and promoting reproducibility. Through illustrative simulation studies motivated by education and behavioral health interventions, we demonstrate how IGR can be used to improve effect estimates compared to benchmark designs in group formation experiments and experiments with interference.