The bivariate Sinh-Elliptical distribution with applications to Birnbaum–Saunders distribution and associated regression and measurement error models
Introduction
Many univariate distributions have been generalized to form bivariate absolutely continuous distributions. These include several common distributions such as uniform, normal, exponential, beta, Weibull and gamma (see Balakrishnan and Lai, 2009). The univariate Sinh-Normal (SN) distribution has received considerable attention in lifetime regression models, especially when the lifetime data involve modeling by the distribution of Birnbaum and Saunders, 1969a, Birnbaum and Saunders, 1969b, in which the scale parameter depends on the covariates. The Birnbaum–Saunders (BS) distribution has applications in a wide range of problems and so does the SN distribution due to its close relationship to the BS distribution. Johnson (1949) mentioned that it is natural and also convenient to build non-normal distributions through transformations of a random variable (rv) that follows a normal distribution. He used the translation method to generate statistical distributions that assume a wide variety of shapes via the transformation where and is a monotone function. These non-normal distributions of have four parameters and , where and correspond to shape parameters while and are location and scale parameters, respectively. The parameters and are assumed to be positive. Based on (1), with , Rieck (1989, p. 13) defined that, if then follows a four-parameter Sinh-Normal (SN) distribution, with its probability density function (pdf) given by where , and denotes the standard normal density function. In this case, the notation is used, which is reduced simply to when in (3); see Rieck and Nedelman (1991). Moreover, when , the r.v. follows the Birnbaum–Saunders (BS) distribution (Birnbaum and Saunders, 1969a) that is usually denoted by , where . This is the reason why the Sinh-Normal is called the log-Birnbaum–Saunders distribution. Some extensions of the SN distributions have been considered in the literature by replacing the normal distribution by a symmetric or asymmetric distribution, as can be seen in Barros et al. (2008) and Leiva et al. (2010), respectively. Another extension could be obtained by considering the log-linear regression model for the -Birnbaum–Saunders distribution; see Cordeiro and Lemonte (2011).
Even though considerable amount of work has been done on univariate Sinh-Normal and Birnbaum–Saunders distributions and their extensions, bivariate versions of these models have received little attention in the literature. The bivariate BS distribution was presented by Kundu et al. (2010) who also derived several interesting and attractive properties of it by using some known properties of the bivariate normal distribution. Hence, it will be natural to study the log-bivariate BS distribution that could also be called the bivariate Sinh-Normal (BSN) distribution. Only recently, Kundu (2014) introduced a BSN distribution based on the normal distribution, which is obtained through the Gaussian copula property.
Even though normality may be a reasonable model to build the BSN distribution, it may lack robustness in parameter estimation under departures from normality (in the form of heavy tails) and/or outliers (Lange and Sinsheimer, 1993).
In this paper, we derive a bivariate extension of the Sinh-Normal distribution by basing it on the bivariate elliptical distribution in place of bivariate normal distribution, and refer to it as the bivariate Sinh-Elliptical (BSE) distribution.
Before presenting our derivation, let us first recall the definition of the multivariate elliptical distribution (Fang et al., 1990): We say that a -variate random vector has an elliptical distribution, denoted by , where is the location vector and is a positive definite dispersion matrix, if its pdf is given by The function is referred to as the density generator. Alternatively, the elliptical distribution is denoted by , where its characteristic function is . Note that the expectation and the covariance matrix of are and Var, respectively, where with being the first derivative of the function associated with the characteristic function of . For this distribution, the kurtosis measure is , where is the second derivative of . Elliptical distributions include several special cases, such as the normal, Student-, Logistic, and Laplace. For an elaborate discussion on elliptical distributions, one may refer to Cambanis et al. (1981), Muirhead (1982), and Fang et al. (1990). For , with , the joint pdf is given by where and is the correlation coefficient.
The main aim of this paper is to present the BSE distribution, which has its marginal distributions and conditional distributions as univariate Sinh-Elliptical distributions that include Sinh-Normal (Rieck, 1989) and Sinh-Student- (Barros et al., 2008), as particular cases. Also, the special case has been discussed in Leiva et al. (2010). The asymptotic behavior of the method of moments estimators is also discussed, and the relationship to the bivariate BS distribution introduced by Kundu et al. (2010) is pointed out. In addition, the associated regression and linear measurement error models are analyzed.
The rest of the paper is organized as follows. In Section 2, we define the BSE distribution and discuss some of its properties. Section 3 discusses the asymptotic properties of the method of moments estimators of the model parameters. Section 4 discusses the relationship between the bivariate Sinh-Elliptical distribution and the bivariate Birnbaum–Saunders distribution of Kundu et al. (2010). In addition, the associated regression model and linear measurement error model are also discussed as an application of the results developed here. In Section 5, numerical examples using both simulated and real data are presented to illustrate the proposed methodology. Finally, some concluding remarks are made in Section 6.
Section snippets
The multivariate Sinh-Elliptical distribution
Analogous to the multivariate BS distribution of Kundu et al., 2010, Kundu et al., 2013, and the skewed Sinh-Normal distribution of Leiva et al. (2010), we define the multivariate Sinh-Elliptical (SE) distribution by considering the following stochastic representation for : where , with being a positive-definite correlation matrix. Thus, the -dimensional random vector is said to have the multivariate SE distribution
Consistent asymptotically normal estimators of the model parameters
Let be independent observations from the model in (5). In order to provide consistent estimators of the model parameters, we consider the following bivariate random vectors obtained from (6): As mentioned earlier, , has an elliptical distribution . Note that the vector depends on and . We can study the asymptotic properties of the 2×2 covariance matrix of
Bivariate BS models
As mentioned earlier, there is a close relationship between the univariate Sinh-Normal distribution and the Birnbaum–Saunders distribution. Here, we will discuss the relationship between the BSE distribution and the bivariate Birnbaum–Saunders distribution proposed recently by Kundu et al. (2010) based on the normal kernel; see also Díaz-García and Domínguez Molina (2007) for the dependent case.
Let , with and . Then, one can consider the
Numerical application
In this section, we consider three simulation experiments to verify if we can estimate the true parameter values accurately by using the proposed method. This is the first step to ensure that the estimation procedure works satisfactorily. In addition, a real data set is analyzed, illustrating the usefulness of the proposed methodology.
Concluding remarks
This paper provides an extension of the univariate Sinh-Normal distribution and of the bivariate Birnbaum–Saunders distribution based on elliptical distributions. We have pointed out some important properties of this new model, which allow us to study its characterization. The BSE distribution, whose marginals and conditionals are generalized Sinh-Normal distributions, admit different degrees of kurtosis and includes as special cases the univariate and bivariate forms of the Sinh-Normal, the
Acknowledgments
The authors thank two reviewers for their helpful comments which have resulted in an improvement over the earlier version of the manuscript. The authors also acknowledge the partial financial support received from FAPEMIG (APQ-01520-12), FAPESP (2011/06263-8) and CNPq, Brazil (309086/2009-4; 303606/2012-6), and the Natural Sciences and Engineering Research Council of Canada.
References (31)
- et al.
On the theory of elliptically contoured distributions
J. Multivariate Anal.
(1981) - et al.
The -Birnbaum–Saunders distribution: an improved distribution for fatigue life modeling
Comput. Statist. Data Anal.
(2011) - et al.
A new family of life distributions for dependent data: estimation
Comput. Statist. Data Anal.
(2007) - et al.
Bivariate Birnbaum–Saunders distribution and associated inference
J. Multivariate Anal.
(2010) - et al.
Generalized multivariate Birnbaum–Saunders distributions and related inferential issues
J. Multivariate Anal.
(2013) - et al.
Improved maximum likelihood estimation in Birnbaum–Saunders nonlinear regressions
Comput. Statist. Data Anal.
(2009) - et al.
On some characterizations of the -distribution
Statist. Probab. Lett.
(2005) - et al.
Ultrastructural elliptical models
Canad. J. Statist.
(1996) - et al.
Continuous Bivariate Distributions
(2009) - et al.
A new class of survival regression models with heavy-tailed errors: robustness and diagnostics
Lifetime Data Anal.
(2008)