Marginal Effects for Meta-Analysis of Nonlinear Outcomes in Criminology: Toward a Practical Understanding of Treatment Effects

• In Event: Quantitative Applications in Criminal Justice Research

Abstract

Meta-analyses of criminal justice interventions often utilize effect size metrics that are nonlinear in nature. Such effect sizes are generally represented using relative ratios of probabilities, odds, or rates, which standardize treatment effects in ways that permit comparison across studies. One cost to this benefit, however, is that treatment effects are no longer interpreted on the natural metric of the dependent variable. Increasingly, scholars have proposed the use of marginal effects to present the results of nonlinear models on their natural scale, but these methods have rarely been applied to meta-analysis, particularly in the field of criminology. In this study I propose the use of marginal effects for meta-analysis in criminology and criminal justice. I also propose multiple methods to calculate the standard error of the marginal effect, and present the results of Monte Carlo simulations testing the coverage and width of the confidence intervals produced by these methods.

Author

• Kevin Petersen, University of North Carolina at Charlotte