We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Statistics > Applications

Title: Review of several false positive error rate estimates for latent fingerprint examination proposed based on the 2014 Miami Dade Police Department study

Abstract: During the past decade, several studies have been conducted to estimate the false positive error rate (FPR) associated with latent fingerprint examination. The so-called Black-box study by Ulery et al. is regularly used to support the claim that the FPR in fingerprint examination is reasonably low (0.1%). The Ulery et al.'s estimate of the FPR is supported by the results of the extensive study of the overall fingerprint examination process by Langenburg. In 2014, the Miami Dade Police Department (MDPD) Forensic Services Bureau conducted research to study the false positive error rate associated with latent fingerprint examination. They report that approximately 3.0% of latent fingerprint examinations result in a false positive conclusion. Their estimate of the FPR becomes as high as 4.2% when inconclusive decisions are excluded from the calculation. In their 2016 report, the President's Council of Advisors on Science and Technology (PCAST) proposes that the MDPD FPR estimate be used to inform jurors that errors occur at a detectable rate in fingerprint examination; more specifically, they declare that false positives may occur as often as 1 in 18 cases. The large discrepancy between the FPR estimates reported by Ulery et al. and Langenburg on the one hand, and the MDPD on the other hand, causes a great deal of controversy. In this paper, we review the MDPD study and the various error rate calculations that have been proposed to interpret its data. To assess the appropriateness of the different proposed estimates, we develop a model that re-creates the MDPD study. This model allows us to estimate the expected number of false positive conclusions that should be obtained with any proposed FPR and compare this number to the actual number of erroneous identifications observed by MDPD.
Subjects: Applications (stat.AP)
Cite as: arXiv:1809.03910 [stat.AP]
  (or arXiv:1809.03910v2 [stat.AP] for this version)

Submission history

From: Madeline Ausdemore [view email]
[v1] Tue, 11 Sep 2018 14:09:21 GMT (1141kb)
[v2] Mon, 22 Oct 2018 16:10:12 GMT (951kb)

Link back to: arXiv, form interface, contact.