Scale and shape mixtures of matrix variate extended skew normal distributions
Introduction
Matrix variate distributions have a central role in multivariate analysis, and have been useful in many fields such as psychology, physics and economics. This type of distributions have proven useful for describing and modeling repeated measurements in multivariate variables. In this context, the observations are not necessarily independent. As an example, consider blood tests for individuals where in each test, variables are measured. In this situation, a random matrix of size is then observed for each individual. In most cases, however, the focus has been focused on matrix variate elliptical distributions, particularly on the matrix variate normal distribution. In the books of Fang and Zhang [11], Gupta and Nagar [15] and Gupta et al. [17] we can find a comprehensive collection of the most important results on matrix variate elliptical distributions.
The matrix variate normal distribution is one of the most well-known matrix variate elliptical distribution. For example, its quadratic forms are used in the analysis of the multivariate linear models. A random matrix of size is said to have a matrix variate normal distribution with parameters , and and is denoted by , if its probability density function (pdf) is given by where . Here, is a matrix of size corresponding to the mean of , while and are positive definite matrices of size and , respectively, such that is the covariance matrix of vec, where vec denotes the vectorization operator.
Matrix variate elliptical distributions, however, cannot take into account the skewness present in the data, which can lead to incorrect conclusions in our analysis. For this reason, in recent years there has been a growing interest in building more flexible matrix variate distributions so that they can adjust the skewness of the data. For instance, a first skew version of the matrix variate normal distribution is introduced by Chen and Gupta [6]. They proposed properties of matrix variate skew normal distribution. A few years later, Domínguez-Molina et al. [10] introduced the matrix variate closed skew normal distribution and gave two constructions for it; and Harrar and Gupta [18] generalized the results of Chen and Gupta [6] and provided a stochastic representation for the matrix variate skew normal distribution. More recently, Zheng et al. [24] studied the inverse problem of matrix variate skew normal distributions; Zheng et al. [25] obtained moments and quadratic forms of matrix variate skew normal distributions by using the moment generating function of the matrix variate closed skew normal distribution. In the same way, Ning and Gupta [21] introduced the matrix variate extended skew normal (ESN) distribution based on the result of Harrar and Gupta [18]. According to these authors, a random matrix of size is said to have a ESN distribution, denoted by , if its pdf is given by where is a location matrix, is a positive definite matrix, and are positive definite matrices, is the pdf of distribution, , and is the cumulative distribution function (cdf) of the -variate normal distribution with mean vector and covariance matrix and , respectively. The moment generating function (mgf) of is where .
On the other hand, there is relatively little work for mixtures of matrix variate distributions. For example, Díaz-García and Gutiérrez-Jáimez [8] proposed the class of scale mixture of matrix variate Kotz-type distributions; Doğru et al. [9] considered finite mixtures of matrix variate distributions; Gallaugher and McNicholas [13] used some finite mixtures of skewed matrix variate distributions for clustering; Gallaugher and McNicholas [12], [14] developed a total of four skewed matrix variate distributions based on a matrix normal variance-mean mixture. In same way, Gupta and Varga [16] proposed the scale mixture of matrix variate normal (SMN) distributions and provided its pdf as follows: where is the pdf of , is a weight function and is a distribution function on . We denote it by .
The main purpose of this paper is to develop the theory of matrix variate flexible distributions by defining the scale and shape mixtures of ESN distributions (hereinafter SSMESN), presenting some particular cases and studying main properties of them. The remainder of this paper is laid out as follows. In Section 2, we give the definition of the SSMESN distributions and propose some subclasses such as scale mixtures of matrix variate ESN (hereinafter SMESN) distributions. Some basic properties are studied in Section 3. In Section 4, we discuss the different representations of the SSMESN distributions. The matrix variate tail conditional expectation (TCE) measure is defined and, also, it is derived for the SSMESN family in Section 5. In Section 6, we obtain maximum likelihood estimator of some special cases of the SSMESN distributions by using the EM algorithm. In Section 7, a small simulation study is presented for evaluating the performance of the maximum likelihood estimators. Finally, some matrix variate distributions in the SSMESN family are fitted to a real dataset in Section 8.
Section snippets
Scale and shape mixtures of matrix variate ESN distributions
In this section, we develop the SSMESN family of matrix variate distributions defined as follows.
Definition 1 Suppose that and are two random variables with joint distribution and marginal distributions and , respectively. A random matrix of size is said to have a scale and shape mixtures of matrix variate extended skew normal distributions with location matrix , a positive definite matrix , two positive definite matrices and and two vectors and , if
Some basic properties
In this section, we derive the distributions of linear and quadratic forms of SSMESN random matrices. We also study here the marginal distributions of SSMESN random matrices.
Selection and stochastic representations
In this section, we discuss the selection and stochastic representations of a SSMESN distributions. We start by adopting the concept of selection distribution presented in Arellano-Valle et al. [3]. For instance, as was considered by Ning and Gupta [21], the matrix variate ESN distribution becomes a selection distribution. In fact, let and and suppose that they are independent. Consider the selection random matrix defined by . By Arellano-Valle et al. [3],
Matrix variate tail conditional expectation
In this section, assuming the existence of the required moments, we introduce the matrix variate tail conditional expectation (TCE) as a new risk measure and obtain it in an explicit closed form expression for the scale and shape mixtures of matrix variate extended skew normal distributions.
Maximum likelihood estimation
In this section, we use the EM algorithm, which was proposed by Dempster et al. [7], to find the maximum likelihood estimator (mle) of , and in SSMESN distributions when and .
A simulation study
To evaluate the performance of the estimators considered in the previous section for varying samples sizes, we consider the matrix variate skew--normal distribution and generate random samples of size and 100, with 300 replications from the , where , , and .
To generate a random sample from , we consider the selection representation of the matrix variate
An example
In this section, we fit some distributions of the SSMESN family, namely, the matrix variate normal, , skew normal, skew and skew--normal distributions, to Dow-Jones dividends dataset. This dataset consists of sum of dividends of Dow-Jones Industrial Common Stocks (DJ dividends) and the Dow-Jones divisor (DJ divisor), which was adjusted to correct the price index for stock splits and stock dividends, for quarters from 1920–1934; see for more Rappaport and White [22]. The data are presented
Conclusions
In the paper, we have presented the definition of scale and shape mixtures of matrix variate extended skew normal distributions. In this definition, we considered the matrix variate extended skew normal distribution which has been introduced by Ning and Gupta [21]. Our definition, with , leads to the definition of a generalization of scale and shape mixtures of multivariate skew normal distributions that is introduced by Arellano-Valle et al. [4]. We obtained the probability density
Acknowledgments
We thank the Editor, Associate Editor and three anonymous referees for the valuable suggestions and comments that improved this paper.
References (25)
- et al.
On some characterizations of the t-distribution
Statist. Probab. Lett.
(1995) - et al.
Scale and shape mixtures of multivariate skew-normal distributions
J. Multivariate Anal.
(2018) - et al.
Compound and scale mixture of matricvariate and matrix variate Kotz-type distributions
J. Korean Stat. Soc.
(2010) - et al.
Finite mixtures of skewed matrix variate distributions
Pattern Recognit.
(2018) - et al.
Three skewed matrix variate distributions
Statist. Probab. Lett.
(2019) - et al.
The multivariate skew-slash distribution
J. Statist. Plann. Inference
(2006) - et al.
Covariance pattern mixture models for the analysis of multivariate heterogeneous longitudinal data
Ann. Appl. Stat.
(2015) - et al.
A unified view on skewed distributions arising from selections
Canad. J. Statist.
(2006) - et al.
On multivariate folded normal distribution
Sankhya B
(2013) - et al.
Matrix variate skew normal distributions
Statistics
(2005)
Maximum likelihood from incomplete data via the EM algorithm
J. R. Stat. Soc. Ser. B Stat. Methodol.
Finite mixtures of matrix variate t distributions
Gazi Univ. J. Sci.
Cited by (3)
Bayes Estimation for the Mean Matrix of the SSMESN Family of Matrix Variate Distributions
2024, Austrian Journal of StatisticsAn Identity for Expectations and Characteristic Function of Matrix Variate Skew-normal Distribution with Applications to Associated Stochastic Orderings
2023, Communications in Mathematics and Statistics