skip navigation

Add your conference to our Justice Events calendar


Register for Latest Research

Stay Informed
Register with NCJRS to receive NCJRS's biweekly e-newsletter JUSTINFO and additional periodic emails from NCJRS and the NCJRS federal sponsors that highlight the latest research published or sponsored by the Office of Justice Programs.

NCJRS Abstract

The document referenced below is part of the NCJRS Library collection. To conduct further searches of the collection, visit the NCJRS Abstracts Database. See the Obtain Documents page for direction on how to access resources online, via mail, through interlibrary loans, or in a local library.

  NCJ Number: NCJ 240957   Add to Shopping cart   Find in a Library
  Title: Goodness-Of-Fit Tests and Function Estimators for Receiver Operating Characteristic (ROC) Curves: Inference From Perpendicular Distances
  Document URL: PDF 
  Author(s): R. Bradley Patterson
  Date Published: 2012
  Page Count: 284
  Annotation: The uniqueness of this research is its use of differences in a perpendicular direction for goodness-of-fit tests, function estimators, and confidence regions for the receiver operating characteristic (ROC) curve.
  Abstract: The ROC curve assesses the performance of classification methods used to identify observations by type. The ROC curve may indicate the performance of a forensic DNA test that classifies genetic samples as to whether or not they match. In order to produce an ROC curve, a sample of observations with known classes must be available. Often, the true ROC function may be a continuous curve that remains unknown. Statistical methods lead to smooth, parametric estimates of the true ROC function from the jagged empirical one. In order to measure the quality of smooth, parametric ROC function estimates, this research project developed a unique statistical goodness-of-fit test. Motivated by that development, the researchers have also created two original estimators of parametric ROC functions. Statistical inference with the ROC curve has traditionally been based on differences between empirical and parametric curves in the vertical direction. What is new and unique in the current work is the use of differences in a perpendicular direction for goodness-of-fit tests, function estimators, and confidence regions for the ROC curve. Working along directions perpendicular to parametric binormal ROC curve, researchers designed a goodness-of-fit test similar to existing statistics based on the empirical distribution function (EDF) for a single random variable. Through large simulations, this new test has exhibited uniformity of p-values under the null hypothesis, and consistency. The original function estimators developed in this work minimize differences in perpendicular directions between empirical and parametric ROC curves. 21 tables 20 figures, and approximately 70 bibliographic listings
  Main Term(s): Criminology
  Index Term(s): Mathematical models ; Statistical analysis ; Comparative analysis ; DNA fingerprinting ; NIJ final report ; NIJ grant-related documents
  Sponsoring Agency: National Institute of Justice (NIJ)
US Department of Justice
Office of Justice Programs
United States of America
  Grant Number: 2011-CD-BX-0124
  Sale Source: NCJRS Photocopy Services
Box 6000
Rockville, MD 20849-6000
United States of America
  Type: Report (Study/Research)
  Country: United States of America
  Language: English
  To cite this abstract, use the following link:

*A link to the full-text document is provided whenever possible. For documents not available online, a link to the publisher's website is provided. Tell us how you use the NCJRS Library and Abstracts Database - send us your feedback.