A TRANSFORMATION FOR EXTENDING GOODNESS-OF-FIT TESTS TO N-VARIABLE DISTRIBUTIONS,

Abstract

The concept of the ordinary probability integral transformation for use with distrubutions of N random variables is extended. This transformation makes possible the evaluation of fit of N-variable distributions to observational data by employing any appropriate 1-variable statistic. In evaluating a model for goodness of fit, statements as to the probability of wrongly accepting a false hypothesis must be discounted, but stated probabilities for wrongly rejecting a true hypothesis are to be taken at face value. The models considered must be of the continuous type and satisfy certain regularity conditions. The transformation is most profitably employed when inadequate sample size precludes use of total information procedures such as the chi-square test with compact multidimensional domain classes. When the available sample size is inadequate to describe an N-variable distribution yet adequate to describe a distribution of one variable then the proposed method is superior to any total information procedure. (Author)

Document Details

Document Type
Technical Report
Publication Date
Sep 23, 1963
Accession Number
AD0426300

Entities

People

  • D. A. Monheit

Organizations

  • Naval Radiological Defense Laboratory

Tags

DTIC Thesaurus Topics

  • Chi Square Test
  • Data Science
  • Goodness Of Fit Tests
  • Information Science
  • Integrals
  • Mathematics
  • Probability
  • Random Variables
  • Test And Evaluation

Fields of Study

  • Mathematics

Readers

  • Applied Combinatorial Optimization and Logic Circuit Design.
  • Calculus or Mathematical Analysis
  • Computational Modeling and Simulation