THREE-LEVEL ANALYSIS OF SINGLE-CASE EXPERIMENTAL DATA

The focus of this postdoctoral research is multilevel modeling of single-case experimental design (SCED) studies in the area of education research. In an SCED, one case is observed repeatedly across time under different levels of at least 1 independent variable (Onghena & Edgington, 2005). During the last decade, the number of published SCED studies is increasing at an astonishing rate in the area of educational research and therefore appropriate statistical modeling methods are needed to summarize these studies. The synthesis of SCED data is of critical importance as it yields significant contributions to research, practice, and policy (Maggin, O’Keeffe & Johnson, 2011). We aim to optimize and extend the multilevel modeling (MLM) framework to synthesize SCED data as little is known about its performance if specific synthesis characteristics and SCED data characteristics are included. A clear overview of these characteristics is missing in the literature and therefore a systematic review of published meta-analyses of SCED studies in the domain of education is needed. A large dataset will be created which will establish the empirical foundation for further studies by SCED researchers, meta-analysts and methodologists. Another important challenge, given the typically small sample sizes in SCED context, is the estimation of the MLM parameters. We present alternative estimation procedures to the traditional maximum likelihood estimation. Once we succeed in dealing with estimation issues, we further extend the MLM framework by including commonly encountered synthesis and SCED data characteristics. We evaluate how powerful the extended multilevel models are.