Comparison of likelihood divergence estatics for testing hypotheses in familial data

  1. Hobza, Tomas
  2. Molina Peralta, Isabel
  3. Morales González, Domingo
Libro:
XXVI Congreso Nacional de Estadística e Investigación Operativa: Úbeda, 6-9 de noviembre de 2001

Editorial: Jaén : Universidad de Jaén, 2001

ISBN: 84-8439-080-2

Ano de publicación: 2001

Congreso: Congreso Nacional de Estadística e Investigación Operativa (26. 2001. Úbeda)

Tipo: Achega congreso

Resumo

For testing composite hypotheses in parametric models, Rényi test statistics are defined as Rényi divergences between nonrestricted and restricted estimated joint probability density functions. This family of statistics is proposed to test the equality of intraclass correlation coefficients in multivariate normal familial data. When maximum likelihood estimators of parameters are plugged in, the asymptotic distribution of these test statistics under the null hypothesis is obtained. Rényi statistics are compared here with likelihood ratio test statistic (LRT) in terms of sizes and powers of the tests.