Title: A Tree-Based Approach for Exploring Moderation Effects in
Meta-Analysis: The R Package metacart

Authors: Xinru Li, Elise Dusseldorp, Xiaogang Su, Jacqueline J. Meulman

Affiliation: Leiden University, Mathematical Institute; Leiden University,
Institute of Psychology; University of Texas at El Paso, Department of
Mathematical Sciences; Department of Statistics, Stanford University

Abstract:
In meta-analyses with multiple moderators, interactions effects may occur
among moderators; that is, moderators may amplify or attenuate each other.
Such interaction effects can be of interest to identify effective
combination(s) of treatment components and to investigate patient subgroups
for which certain treatment components result in a larger effect. However,
standard meta-regression methods often lack power to identify interactions,
especially when the number of moderators is large. To remedy this
disadvantage, we propose to use tree-based methods as an alternative
methodology, since they are dealing with convoluted interaction effects in
high-dimensional data of mixed type, and because their results are easily
interpretable. Our recently proposed method, termed meta-CART (Li et al. 2017,
2018a), integrates classification and regression trees into the meta-analysis
framework to explore interactions among moderators. The approach is
implemented in the R-package metacart (Li et al. 2018b, available on CRAN);
the package can fit both fixed- and random-effects models, and can handle
dichotomous, categorical, ordinal and continuous moderators. In this
presentation, we will describe the algorithm underlying the package, and show
the results of an application to a real-world data set.

References:
Li, X., Dusseldorp, E. and Meulman, J. J. (2017), Meta-CART: A tool to
identify interactions between moderators in meta-analysis, British Journal of
Mathematical and Statistical Psychology, 70 (1), 118-136.  

Li, X., Dusseldorp, E., and Meulman, J. (2018), A  flexible approach to
identify interaction effects between moderators in meta-analysis, Research
Synthesis Methods. Accepted. 

Li, X., Dusseldorp, E., Su, X., and Meulman, J. (2018), Multiple moderator
meta-analysis with R-package metacart. Submitted.