Measures of relative model fit

Alan Agresti, Brian Caffo

Research output: Contribution to journalArticle

Abstract

Most software for statistical model fitting reports the value of the maximized log likelihood function, but the numerical value is difficult to interpret because of its log scale, its dependence on the sample size, and the possible omission of constants. A sample-size-scaled version of the likelihood function summarizes the model fit. A related index uses the mean of the contributions to the likelihood function. The ratio or difference of either index for two models is a summary measure of relative model fit. We discuss these measures and briefly consider interval estimation for them.

Original languageEnglish (US)
Pages (from-to)127-136
Number of pages10
JournalComputational Statistics and Data Analysis
Volume39
Issue number2
DOIs
StatePublished - Apr 28 2002
Externally publishedYes

Keywords

  • AIC
  • Deviance
  • Discrete data
  • Dissimilarity
  • Likelihood function
  • Non-nested models

ASJC Scopus subject areas

  • Statistics and Probability
  • Computational Mathematics
  • Computational Theory and Mathematics
  • Applied Mathematics

Fingerprint Dive into the research topics of 'Measures of relative model fit'. Together they form a unique fingerprint.

Cite this