Effect Estimation in Randomized Single-Case Experimental Designs: The Influence of Undetected Extraneous Variables

• In Event: Effect Estimation for Causality and Prediction: Multilevel and Single-case Designs, Machine Learning, and Distributional Assumptions

Abstract

A challenge in planning and analyzing single-case experimental designs is that treatment effect estimates may be biased by the effects of extraneous variables (e.g., changes in the school environment that impact a child’s behavior, but that are not part of the intervention). We developed a simulation study to estimate the impact of extraneous variable effects on a variety of single-case design effect sizes (i.e., raw score mean difference, within-case standardized mean difference, and percent of goal obtained) for designs that incorporated different forms of randomization (intervention start-point randomization and intervention-order randomization), and across a variety of conditions that vary in series length, number of participants, and effect size. Implications for the design and analysis will be discussed.

Authors

• John M. Ferron, University of South Florida

• Moses Muchiri Mohamed, University of South Florida

• Joel R. Levin, University of Arizona

• Tom Kratochwill