Estimating effect size with respect to variance in baseline to treatment phases of single-case experimental designs: A Bayesian simulation study

Abstract

Abstract The current study examined the relation between the ratio of baseline to treatment sessions and how differences in this ratio can influence estimation of treatment effect size from temporally adjacent baseline and treatment phases of any single-case experimental design (SCED). The current study describes how Bayesian statistical analyses can be used to aggregate treatment outcomes across subjects to meta-analyze SCED data. One-third of all A versus B comparisons (based upon simulated average values) did have a 10% or more bias, with the vast majority of the bias being substantially fewer data points in baseline compared to treatment sessions. SCEDs require relatively steady state responding; thus researchers may run relatively more B sessions compared to A sessions in the course of visually inspecting graphically depicted data. When the standard deviation for the number of A sessions was approximately twice as large or more than the B phase standard deviation, the degree of AB sessions ratio bias decreased substantially. SCED practitioners can use results of the current study to determine the potential benefits of running additional baseline or treatment sessions.