Multilevel Modeling of Single-Case Experimental Data: Handling data and design complexities

In single-case or single-subject designs (SSED), individual cases are measured repeatedly under different conditions, in order to assess the effect of the condition. Recently, multilevel models were proposed to combine the results of SSED studies, resulting in more general or detailed conclusions. The purpose of the proposed research is to empirically investigate the multilevel approach, using both real data and simulation studies. The research will entail several studies designed to address major complications encountered when synthesizing results from SSED research (see also projects 3H110617, 3H150687, 3H150316 and 3H250079). This specific research project focuses on the problems encountered when applying the multilevel model for the meta-analysis of SSED count data. Simulation studies are set up to analyze how robust the model is against misspecifications: what happens when the basic continuous model is applied to non-normal discrete data? A second question is how to apply the model for the statistical meta-analysis of count data: how can count data be standardized and how can the meta-analysis combine primary studies which report both continuous and count data? The results of the simulation studies will also be illustrated with empirical datasets. The intent of this dissertation is twofold. On the one hand, we empirically validate the basic three-level model and several extensions to it using large simulation studies and giving empirical illustrations. On the other hand, we want inform applied applied SSED researchers about the value of multilevel modeling of SSEDs and how to use these models. As a consequence, this dissertation is informative for methodologists, research analysts and synthesists, but also for applied SSED researchers.status: publishe