Talk 2

X Workshop of Probability and Statistics group

CIDMA- University of Aveiro

Speaker: Manuela Souto de Miranda - CIDMA, University of Aveiro

Title of the talk: Robust Estimates with Hurdle Models

Abstract: Hurdle models are mixture models that can help statisticians in modelling data sets with an excess of zeros. Such data sets occur very often in counting processes, particularly, associated with health studies or with economic analyzes. The main characteristic that distinguishes the hurdle model from others that also deal with too many zeros, is the assumption that the random process which generates the strictly positive counts is not the same that generates the occurrence of the zeros. It is easy to find real examples that suggest an adequate hurdle modelling, some of them already studied in the statistical literature. They might be the case of the number of falls of patients with a specific disease or the number of visits to a medical office; or, the number of missing monthly credit payments to a bank. The estimation of the model parameters is generally supported by the maximum likelihood estimators, but robust estimators are not so sensitive to deviations from assumptions, nor to the stability of the optimizing solutions. Nevertheless, robust estimators for the hurdle model parameters are not so deeply studied. This talk will be focused on robust estimation for the hurdle Poisson model, remembering the model and presenting some developments in robust alternative estimators.

Manuela Souto de Miranda, has been collaborating with CIDMA since her retirement, after more than thirty years of teaching and research in the Mathematics Department, University of Aveiro; she also carried out several management positions in the University of Aveiro. Her Master’s degree in Applied Mathematics was obtained in 1988 by Instituto Superior Técnico, University of Lisbon, Portugal, and her PhD was completed in 1995 at the University of Aveiro, Portugal. Currently, she informally collaborates with other research centers and she is affiliated on scientific societies like the International Association for Statistical Computing, and the Portuguese Statistic Society, where she also performed managing roles in both associations.

Orcid: 0000-0002-5703-5082.

Page updated

Google Sites

Report abuse