Efficacy of Bayesian Rate Ratio for Estimating Effect Size of Single-Case Experimental Designs

• In Event: Statistical Inference and Bayesian Estimation

Abstract

Statistical methods used in Single Case Experimental Designs (SCED) are mostly suitable for analyzing continuous data. However, count outcome is more prevalent in SCED. The Bayesian Rate Ratio effect size (BRR) is a within subject-effect size for autocorrelated count data in SCED analysis. The current paper examines the efficacy of BRR as an effect size using simulation under various conditions of phase lengths, autocorrelation values, and standardized mean difference effect size. Preliminary results of a 2×2×2 full factorial design with 5 replications show the Root Mean Squared Error (RMSE) of the posterior mean of effect size is lower for longer phase length, low autocorrelation value, and smaller mean difference between the phases.

Authors

• Pragya Shrestha, University of Delaware

• Prathiba Natesan Batley, University of Louisville

• Ratna Nandakumar, University of Delaware