Title: Three-level meta-analytic models to combine several types of single-case designs Authors: Moeyaert, Mariola ×Van den Noortgate, Wim # Issue Date: 22-Aug-2014 Conference: International Conference on Computational Statistics edition:21 location:Genève, Switzerland date:19-22 August 2014 Abstract: As the number of published single-case studies is increasing at an astonishing rate during the last decade, there is a need to optimize the statistical techniques to quantify the research findings in an objective way. A single-case experimental study is …a designed experiment in which one case (i.e., a unit which can be one subject or a small group of subjects) is observed repeatedly during a certain period under different levels (\emph{treatments}) of at least one independent variable.'' When using data from multiple single-cases, a three-level structure becomes visible: measurement occasions are nested within subjects and subjects in turn are nested within studies. The raw single-case data can be synthesized across subjects and across studies using a multilevel model which is an extension of the regression approach. The purpose of the study is to further extend the multilevel meta-analysis of single-case data by including complex single-case designs such as alternating treatment designs and ABAB reversal designs in addition to the multiple-baseline across participants design (which are the most popular single-case designs). We suggest two univariate models and one multivariate multilevel model and illustrate the proposed methods using a published meta-analysis of single-cases including the three types of single-case designs. Publication status: published KU Leuven publication type: IMa Appears in Collections: Methodology of Educational Sciences