ABSTRACT
This chapter reviews statistical models to quantify the effects of factors influencing the accuracy of eyewitness identification in controlled experiments as well as to explore methods for analyzing the results from these experiments, using statistical models and intuitive displays of the effects of these factors. For example, although the receiver operating characteristic (ROC) curve has been used for decades in statistical quality control, diagnostic medicine, and many other fields where methods or techniques are being compared, the ROC curves using data from eyewitness identification experiments are constructed using the experimental participant's expressed confidence level (ECL) in the identification, which can be affected by error and variation. We present alternative statistical approaches, some of which have been used in similar scenarios (e.g., comparing medical diagnostic imaging modalities) with the aim of developing more powerful analyses to better quantify the effects of variables (including or modifying the ECL) influencing the accuracy of EWID procedures. These statistical tools may offer powerful ways of identifying factors that affect EWID accuracy, beyond the conventional tools of diagnosticity ratios and ROC. This chapter serves to provide information on existing and viable statistical methods for analyzing EWID experiments. Whichever technique is used, proper characterization of the uncertainties associated with inferences must be calculated.
