Advancing The Use Of Receiver Operating Characteristic (ROC) Curves To Measure Performance Of Forensic Methods

The aim of this study is to improve the use of Receiver Operating Characteristic (ROC) Curves to measure the effectiveness of different forensic methods. The Principal Investigator, R. Bradley Patterson, will plot true positive rates versus false positive rates for each forensic method. With these normalized measures, we may compare the methods objectively by comparing their ROC curves. Moreover, ROC curves provide the full range of accuracy rates and many additional measures of effectiveness which make the comparisons comprehensive. In order to develop a parametric ROC curve that offers statistical efficiency, this study will develop a goodness-of-fit test based on a family of statistics originally designed to test the fit of similar functions. The study will focus on methods that evaluate glass fragments analyzing data from a publically available dataset provided by the Journal of the Royal Statistical Society that contains measurements of the elemental composition of 310 glass fragments.

This research will give forensic scientists the ability to describe effectiveness with the standard scales of false positive and false negative rates, allowing for the comparison of forensic methods that may use very different techniques for evaluating the same evidence. Thus, this research will give criminal justice agencies the ability to rely on ROC curves for complete and objective comparisons when considering new methods of evaluating evidence.
ca/ncf