Multivariate log-Birnbaum–Saunders regression models

ABSTRACT

In this paper, we present a multivariate version of the skewed log-Birnbaum–Saunders regression model. This new family of distributions holds good properties such as marginal variables following univariate skewed log-Birnbaum–Saunders distributions, besides presenting the usual log-Birnbaum–Saunders distribution as a particular case. Furthermore, the model parameters are estimated through maximum-likelihood methods, a closed-form expression for the Fisher’s information matrix is presented, and testing hypothesis for model parameters is performed. Two real datasets are analyzed and results are discussed.

KEYWORDS:

• log-linear Birnbaum–Saunders model
• sinh-normal distribution
• skewed model.

MATHEMATICS SUBJECT CLASSIFICATION:

• 62H10
• 62J05

Acknowledgment

We are very thankful for the very constructive comments made by the two reviewers.