Comparing the percentage of non-overlapping data approach and the hierarchical linear modeling approach for synthesizing single-case studies in autism research

Abstract

We examined the performance of two approaches for synthesizing single-case experimental data: the percentage of non-overlapping data (PND) approach and the hierarchical linear modeling (HLM) approach. The comparison was performed by analyzing an empirical dataset on behavioral interventions for reducing challenging behavior in persons with autism by means of the two approaches. We compared the findings of both approaches for analyzing the outcomes of the behavioral interventions as well as for identifying moderating variables. With respect to the analysis of the interventions’ outcomes, similar positive results were found based on both approaches. With respect to the moderating variables, Functional analysis/assessment and Availability of follow up data were found to be statistically significant moderators by means of the PND as well as the HLM approach. The variables Intervention type, Availability of generalization attempts, Design type, and Availability of inter-rater reliability data were also found to be statistically significant moderators by means of the PND approach. The PND approach seems overly liberal in identifying statistically significant predictors in comparison to the HLM approach.