Conference Prof. Tomáš Hobza

5 junio, 2017
12:00 pma2:00 pm

Tittle: A Wald-type test statistic based on robust modified median estimator in logistic regression models

Speaker:  Tomáš Hobza (Czech Technical University in Prague)

Date: 05/06/2017, 12:00h

Location: Sala de Seminarios, Edificio Torretamarit

Abstract

This contribution deals with robust estimation and hypothesis testing in logistic regression models. It introduces a modified median estimator of the underlying parameters of these models based on statistically smoothed binary responses and gives its asymptotic distribution.

Using the modified median estimator, a Wald-type test statistic for testing linear hypotheses in the logistic regression model is also considered and its asymptotic distribution is obtained.

Sensitivity to contaminations and leverage points is studied by simulations and compared in this manner with the sensitivity of some robust estimators and tests previously introduced to the logistic regression. The new estimator and test based on it appears to be more robust for larger sample sizes and higher levels of contamination.

Key words: Logistic regression, Median estimator, Robustness, Wald-type tets