A Generalization of the Skew-Normal Distribution: The Beta Skew-Normal

TL;DR

This paper introduces the Beta skew-normal distribution, a new family generalizing the skew-normal, derived from order statistics, with properties, special cases, and methods for simulation, expanding the understanding of skew-normal generalizations.

Contribution

The paper proposes the Beta skew-normal distribution, a novel generalization of the skew-normal, and explores its properties, special cases, and relationships with existing distributions.

Findings

The Beta skew-normal contains the Beta half-normal as a limiting case.

Derived the moment generating function and moments of the BSN.

Presented methods for simulating the BSN distribution.

Abstract

The aim of this article is to introduce a new family of distributions, which generalizes the skew normal distribution (SN). This new family, called Beta skew-normal (BSN), arises naturally when we consider the distributions of order statistics of the SN. The BSN can also be obtained as a special case of the Beta generated distribution (Jones (2004)). In this work we pay attention to three other generalizations of the SN distribution: the Balakrishnan skew-normal (SNB) (Balakrishnan (2002), as a discussant of Arnold and Beaver (2002), Gupta and Gupta (2004), Sharafi and Behboodian (2008)), the generalized Balakrishnan skew-normal (GBSN) (Yadegari et al. (2008)) and a two parameter generalization of the Balakrishnan skew-normal (TBSN) (Bahrami et al. (2009)). The above three extensions are related to the Beta skew-normal distribution for particular values of the parameters. The paper is organized as follows: after describing briefly, in Section 2, the skew-normal distribution, its generalizations and listing their most important properties, in Section 3 we present some generalizations of the Beta distribution. In the last Section we define the Beta skew-normal distribution, we present its properties and some special cases. In particular the BSN contains the Beta half-normal distribution (Pescim et al. (2010)) as limiting case. Besides, we investigate its shape properties. We derive its moment generating function and we also compute numerically the first moment, the variance, the skewness and the kurtosis. We present two different methods which allow to simulate a BSN distribution. We explore its relationships with the other generalizations of the SN. We give some results concerning the SNB distribution. In particular we derive the exact distributions of the largest order statistic from SNB with parameters m and 1 and the shortest order statistic SNB with parameters m and -1.