'On Detecting Outliers in Mixed Populations'.

Abstract

In this report we introduce an operational methodology for detecting outliers in data which is a mixture of events from a variety of sources. The only assumption required is that the data contain no previous nuclear events. Thus ground truth data is not required. The method models the data as a mixture of two or more event types. It then develops a test statistic based on a modified likelihood ratio and the bootstrap T. test for outliers. The calculation of this statistic is accomplished by making use of clustering methods to initialize the EM algorithm which is then used to obtain the required maximum likelihood estimates.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Feb 01, 1996
Accession Number
ADA308972

Entities

People

  • H. L. Gray
  • S. R. Sain
  • W. A. Woodward
  • W. H. Frawley

Organizations

  • Southern Methodist University

Tags

Communities of Interest

  • Energy and Power Technologies

DTIC Thesaurus Topics

  • Algorithms
  • Anomaly Detection
  • Atmospheric Sciences
  • Change Detection
  • Clustering
  • Contracts
  • Data Science
  • Databases
  • Detection
  • Earth Sciences
  • Geography
  • Geophysics
  • Information Science
  • Planetary Sciences
  • Statistical Analysis
  • Statistical Distributions
  • Surveys

Fields of Study

  • Mathematics

Readers

  • Aerospace Test and Evaluation
  • Statistical inference.