In recent years, the forensic community has pushed to increase the scientific basis of forensic evidence, which has included proficiency testing for fingerprint analysts. We used proficiency testing data collected by Collaborative Testing Services in which 431 fingerprint analysts were asked to identify the source of latent prints. The data were analysed using a Rasch model with a Bayesian estimation approach. Although these data provide valuable information about the relative proficiency of the examiners and the relative difficulty of the questions, it does not necessarily extrapolate onto general performance of examiners or difficulty in casework, which we show through sensitivity analysis and simulation. We show that a Bayesian Item Response Theory (IRT) analysis provides a deeper understanding of analysts’ proficiency and question difficulty than other forms of analysis. A large-scale adoption of IRT in this area would provide both more precise estimates of proficiency and quantitative evidence for the relative difficulty of different questions.
Proficiency Testing of Forensic Fingerprint Examiners with Bayesian Item Response Theory

Journal: Law, Probability and Risk
Published: 2018
Primary Author: Amanda S. Luby
Secondary Authors: Joseph B. Kadane
Type: Publication
Research Area: Latent Print
Related Resources
Latent Print Quality in Blind Proficiency Testing: Using Quality Metrics to Examine Laboratory Performance
Presented at American Association of Forensic Sciences (AAFS) 2021
The Implementation of a Blind Quality Control Program in a Forensic Laboratory
Presented at American Association of Forensic Sciences (AAFS) 2021
A Field Analysis of Laboratory Case Processing: Latent Print Comparison and Examiner Conclusions
Presented at American Association of Forensic Sciences (AAFS)
Psychometrics for Forensic Fingerprint Comparisons
Forensic science often involves the evaluation of crime-scene evidence to determine whether it matches a known-source sample, such as whether a fingerprint or DNA was left by a suspect or…