Dados do Trabalho

Título

A Robust multivariate Birnbaum-Saunders regression models

Resumo

This work focus on multivariate Birnbaum-Saunders (BS) regression model that { can be used} in survival analysis to analyze correlated log-lifetimes of two or more units. This multivariate BS regression models is studied through the use of a generalization multivariate of the sinh-normal (SN) distribution, which is built from
the multivariate mixture scale of normal (SMN) distribution. The marginal and conditional linear regression models of the resulting multivariate BS linear regression model are generalizations of the Birnbaum–Saunders linear regression models (Rieck and Nedelman, 1991), which has been used effectively for modelling lifetime data and reliability problems. We exploit the nice hierarchical representation of the generalized multivariate SN to propose a fast and accurate EM algorithm for computing the maximum likelihood (ML) estimates of the model parameters. Hypothesis testing is also performed by the use of the asymptotic normality of the ML estimators.

Finally, the results of simulation studies as well as an application to a real data set are displayed, where also is included robustness feature of the estimation procedure developed here.

Palavras-chave

Birnbaum-Saunders distribution; Sinh-normal distribution; EM algorithm; Maximum likelihood method;
Robust estimation; Scale mixtures of normal distributions; Asymptotic normality; Hypothesis testing.