Individual Participant Data Meta-Analysis of Single-Case Experimental Design Data: A Monte Carlo Simulation Study

• In Event: AERA Poster Session Wednesday 3:45pm
  In Poster Session: Systematic Review and Meta-Analysis: Methods, Contexts, and Impacts

Abstract

Estimating parameters in two-stage individual participant data (IPD) meta-analysis (MA) of single-case experimental designs (SCEDs) is challenging. Despite the advantages of Bayesian estimation, no methodological studies have evaluated its performance in IPD MA of SCEDs. We conducted a Monte Carlo simulation comparing Bayesian and restricted maximum likelihood (REML) methods for estimating the overall intervention effect and its variance components. Both Bayesian and REML estimations produced unbiased estimates of the overall intervention effect, with REML preferred for detecting the effect. Bayesian methods generally outperformed in estimating variance components, especially when using priors with more information. Neither Bayesian nor REML estimation consistently yielded unbiased standard error estimates for both variance components. The scientific significance and limitations are discussed.

Authors

• Yukang Xue, University at Albany - SUNY

• Mariola Moeyaert, University at Albany - SUNY