LIKELIHOOD RATIO AND SCORE TESTS APPLIED TO THE LOGISTIC REGRESSION MODEL

Enrico Antonio COLOSI1

Glaura da Conceição FRANCO1

Bráulio Roberto Gonçalves Marinho COUTO[1]

§    ABSTRACT: Hypotheses tests for building logistic regression models are based on the asymptotic properties of the maximum likelihood estimators. Wald’s score and likelihood ratio tests are used according to the client’s taste and computer software availability. Wald’s test may show aberrant results ( Hauck & Donner, 1977) in addition to the fact that it is not invariant to 1-1 transformations of the parameters. In this work, small sample properties of the score and likelihood ratio tests are compared in real life situations.

§    KEYWORDS: Maximum likelihood; score test; likelihood ratio test; power function; asymptotic properties.

 



[1] Departamento de Estatística – ICEx -Universidade Federal de Minas Gerais - UFMG - 30161-970 - Belo Horizonte - MG - Brasil.