Meta-CART: An integration of classi cation and regression trees into meta-analysis
Meta-CART: integrate classification and regression tree into meta-analysis
Meta-analysis is an important tool to synthesize results from multiple studies in a systematic way. Interaction effects play a central role in assessing conditions under which the relationship between study features and effect size (the outcome variable) changes in strength and/or direction. However, within the framework of meta-analysis, interaction effects between moderators are barely investigated due to the lack of theories for confirmatory studies and methods with enough power for exploratory studies. To detect interaction effects in exploratory studies, a new approach named “meta-CART” introduced Classification and Regression Trees (CART) in the field of meta-analytic data to identify interactions. The current version of meta-CART has several shortcomings: 1) when applying CART, the sample sizes of studies are not taken into account; 2) the effect size is dichotomized around the median value; 3) the method is a stepwise approach. In this PhD project, we will propose new extensions for meta-CART to overcome its shortcomings and to improve its performance. Furthermore, Monte Carlo simulation studies will be carried out to valuate the performance of (extended) meta-CART. As a result, software will be developed for researchers to apply meta-CART for interaction detection in real-world data.
Prof. Jacqueline J. Meulman (Leiden University)
Dr. Elise Dusseldorp (Leiden University)
1 November 2014 – 27 February 2020