Journal of Statistical Computation and Simulation, vol.94, no.2, pp.400-423, 2024 (SCI-Expanded)
The motivation of this study is to obtain more accurate results in meta-regression by modeling the covariates used in explaining the heterogeneity between studies included in the meta-analysis. For this purpose, the CERES plots that are already used in regression models are adapted to meta-regression to better visualize the form of the included covariates. To demonstrate the success of the CERES plot, two types of simulation data are derived based on the form of dependency between the two covariates. Also, the real data is considered. It has been found that CERES plots produce a good performance in deciding on the proper form of the variables for independent and dependent scenarios as well as real data. On the other hand, the regression parameters that represent the magnitude of the impact of the covariates on the effect size are estimated more accurately.