Exploring the Application of Randomization-Based Inference in Experimental and Quasi-Experimental Designs

Posted: May 30, 2017

• Grace Flores
• Theodore Moore
• Sarah Young

Abstract

Randomization has long been recognized as the gold standard for establishing causal relationships in experimental research. The fundamental principle of randomization-based inference (RBI) rests on the idea that the act of random assignment provides a known probability distribution for test statistics under the null hypothesis of no treatment effect. While traditional RBI methods have been predominantly applied in randomized controlled trials, their potential for extension to quasi-experimental designs remains largely unexplored. This paper addresses this gap by developing a comprehensive framework that extends RBI principles to both experimental and quasi-experimental contexts, thereby creating a unified approach to causal inference. The novelty of our approach lies in reconceptualizing randomization not merely as a design feature but as the foundational element for statistical inference across diverse research settings. We propose that the logical structure of randomization tests can be adapted to situations where true randomization is absent by constructing reference distributions based on plausible randomization mechanisms. This represents a significant departure from conventional approaches that typically rely on model-based assumptions or instrumental variables in quasi-experimental settings.