Testing linear hypotheses in logistic regression analysis with complex sample survey data based on phi-divergence measures

Abstract

In this paper a family of Wald-type test statistics for linear hypotheses in the logistic regression model with complex sample survey data is introduced and its properties are explored. The family of tests considered is based on the pseudo minimum phi-divergence estimator that contains, as a particular case, the pseudo maximum likelihood estimator. We obtain the asymptotic distribution and through a simulation study it is shown that some Wald-type tests present much more stable levels than the classical one for high and moderate values of the intra-cluster correlation parameter.

Keywords:

• Logistic regression models
• sample survey data
• pseudo maximum likelihood estimator
• pseudo minimum phi-divergence estimators
• Wald-type tests

Additional information

Funding

This research is partially supported by Grant PGC2018-095194-B-I00 and Grant FPU16/03104 from Ministerio de Ciencia, Innovacion y Universidades and Ministerio de Educacion, Cultura y Deporte (Spain).