Internal meta-analysis for Monte Carlo simulations

Monte Carlo (MC) simulations are one of the dominant approaches to compare statistical methods. To date, there is no standard procedure for MC simulations. Although internally valid, they exhibit a certain degree of arbitrariness through the various choices that researchers make. In this paper, we propose the use of an internal meta-analysis for MC simulations to allow a standardized analysis, synthesis and presentation of MC simulation results in a transparent manner. The use of an internal meta-analysis allows (i) a much more standardized procedure and (ii) comprehensive analysis of a large variety and number of simulations. To exemplify the procedure, we conduct an extensive set of simulations to compare the empirical performance of three different estimators of the generalized stochastic frontier panel data model. Besides contributing to the literature on efficiency analysis by improving the understanding of the merits of the three different estimators, we demonstrate the applicability and usefulness of internal meta-analysis for MC simulations in general.