Crime Mapping

Abstract

Overview Simulation models are becoming increasingly popular in criminology research. In order for researchers to have confidence in the results of simulation studies, it is essential to make sure that the models are properly evaluated. Calibration is a major element to this evaluation and refers to the estimation and adjustment of model parameters to improve the agreement between model output and a data set. This entry will discuss some different methods of model calibration as well as related quantitative methods of error assessment. Introduction Simulation models are becoming increasingly popular in criminology research, both as tools for exploring real-world crime patterns (explanatory models) and for experimenting with underlying theory (conceptual models). In order for researchers to have confidence in the results of simulation studies, it is essential to make sure that the models are properly evaluated. The process of evaluation is commonly divided into three activities, defined by Rykiel (1996) as follows: • Verification is a demonstration that the modeling formalism is correct. • Calibration is the estimation and adjustment of model parameters and constants to improve the agreement between model output and a data set. • Validation is a demonstration that a model within its domain of applicability possesses a satisfactory range of accuracy consistent with the intended application of the model. It is possible to calibrate a model quantitatively by assessing error using statistics or qualitatively by manually comparing model results and field data. Qualitative approaches are particularly well suited to spatial models where the researcher is able to compare maps. However, for complicated models with many parameters, a qualitative approach to calibration will probably be very time consuming and unlikely to reveal the optimal model configuration (which is the ultimate aim of calibration). Therefore, this article will focus on quantitative methods of error assessment (rather than relying on human objectivity) and automatic calibration routines that are able to explore a model's parameter space and estimate error without human intervention. Also, in the context of environmental crim-inology, it is much more important to accurately reflect field conditions when working with explanatory models-i.e., those that simulate real-world conditions-so the discussion will focus on these in particular. Conceptual models, G. Bruinsma, D. Weisburd (eds.), Encyclopedia of Criminology and Criminal Justice,