Eyewitness identification discriminability: ROC analysis versus logistic regression |
| |
| Affiliation: | 1. University of Oklahoma, United States;2. The University of Alabama, Huntsville, United States;1. Department of Psychology, University of California, San Diego, United States;2. Department of Psychology, Royal Holloway, University of London, United Kingdom;1. School of Education, Bar-Ilan University, Ramat-Gan, Israel;2. Department of Psychology and Leslie and Susan Gonda (Goldschmied) Multidisciplinary Brain Research Centre, Bar-Ilan University, Ramat-Gan, Israel;1. University of Oklahoma, Norman, OK 73019-2007, United States;2. Texas A&M University-Commerce, Commerce, TX 75429, United States;1. Department of Psychology, Royal Holloway, University of London, United Kingdom;2. Department of Psychology, University of California, Riverside, United States;3. Department of Psychology, University of California, San Diego, United States |
| |
| Abstract: | To reach conclusions regarding the respective accuracy of two conditions, eyewitness researchers evaluate correct and false identification rates computed across participants. Two approaches typically are employed. One approach relies on ratio-based probative value measures; but Wixted and Mickes (2012) and Gronlund, Wixted, and Mickes (2014) showed that these measures fail to disentangle an assessment of accuracy (i.e., discriminability between guilty and innocent suspects) from response bias (i.e., a willingness to make a response). Our focus is on a second approach, logistic regression analyses of the correct and of the false identification rates. Logistic regression also fails to disentangle discriminability from bias. Therefore, it only can denote the most accurate condition in limited circumstances. The best approach for reaching the proper conclusion regarding which condition is most accurate is to use receiver operator characteristic (ROC) analysis. Simulated ROC data illustrate the problem with a reliance on logistic regression to assess accuracy. |
| |
| Keywords: | Eyewitness identification ROC analysis Signal detection theory Logistic regression Probative value measures |
| 本文献已被 ScienceDirect 等数据库收录！ |
|
