APPLICATION AND VALIDATION TEST OF A GENERAL PREDICTION METHOD FOR HANDSCRIPTION ERROR RATE.

Abstract

Application and validation testing of a previously developed General Prediction Method (GPM) for estimating handscribed source data error rates is described. The GPM application procedure is prescribed, and its application to five operational source data systems at Tinker AFB, Oklahoma, is related. Validation testing of the GPM first required measurement of actual error levels in each of the five data systems in order to develop the confidence intervals necessary. Three aspects of GPM validation were considered. The first involved a confidence interval around the overall sample error ratio for each data system, where the interval was computed on the basis of the overall sample size and an assumption of normality. The second related the overall sample error ratio to the standard error of the estimate. And the third utilized statistical characteristics of each data system as determined by daily samples, to establish validation criteria. In the first approach to validation, none of the five GPM estimates were within the overall sample confidence intervals. In the second approach, four or five of the observed data system error ratios were within the expected limits of accuracy of the GPM (represented by the standard error of the estimate). In the third approach, which showed that the statistical assumptions inherent in the first approach were inappropriate, three of the five data systems were within validation criteria established by the statistical characteristics of each data system. (Author)

Document Details

Document Type
Technical Report
Publication Date
Jul 01, 1968
Accession Number
AD0675217

Entities

People

  • Bruce N. Mcarthur
  • Richard L. Hawley

Organizations

  • FMC Corporation

Tags

DTIC Thesaurus Topics

  • Accuracy
  • Data Science
  • Errors
  • Information Science
  • Intervals
  • Measurement
  • Normality
  • Oklahoma
  • Standards
  • Validation

Readers

  • Approximation Theory.
  • Atmospheric Remote Sensing.
  • Statistical inference.