On The Robustness of Epsilon Skew Extension for Burr III Distribution on Real Line
Extracted source text · arXiv 1705.09988
%-------------------------------------------------------------------------------
% THIS FILE IS AN EXAMPLE ON HOW TO USE THE "REVSTAT.STY" PACKAGE...
% last modified: 2002/11/23
%-------------------------------------------------------------------------------
% REVSTAT - STATISTICAL JOURNAL
% Instituto Nacional de Estatística - Portugal
% http://www.ine.pt/revstat.html
% liliana.martins@ine.pt
%-------------------------------------------------------------------------------
% The following 10 packages are pre-loaded:
%
% graphicx latexsym enumerate amsmath amsfonts amssymb
% amsthm amsopn amstext amscd
%-------------------------------------------------------------------------------
\documentclass[twoside,leqno,11pt]{article}
\usepackage{revstat}
\usepackage{epstopdf}
\usepackage{amsmath}
\usepackage{graphicx}
\usepackage{subcaption}
\begin{document}
%-------------------------------------------------------------------------------
% Title of the article...
%-------------------------------------------------------------------------------
\title{On The Robustness of Epsilon Skew Extension for Burr III Distribution on Real Line
\footnote{The opinions expressed in this text are those of the authors and do not necessarily reflect the views of any organization.}}
\renewcommand{\titleheading}
{Using the ``revstat.sty" Package} % --- on odd running heads
%-------------------------------------------------------------------------------
% Author(s)' names and addresses...
%-------------------------------------------------------------------------------
\author{\authoraddress{Mehmet Niyazi \c{C}ankaya}
{Department of Statistics, Faculty of Arts and Science
U\c{s}ak University,\\
U\c{s}ak
\ (mehmet.cankaya@usak.edu.tr)}
\\
\authoraddress{Abdullah Yal\c{c}{\i}nkaya}
{Department of Statistics, Faculty of Science
Ankara University,\\
\ (ayalcinkaya@ankara.edu.tr)}
\\
\authoraddress{\"{O}mer Alt{\i}nda\v{g}}
{Department of Statistics, Faculty of Arts and Science
B\.{i}lec\.{i}k \c{S}eyh Edebal\.{i} University,\\
\ (ayalcinkaya@ankara.edu.tr)}
\\
\authoraddress{Olcay Arslan}
{Department of Statistics, Faculty of Science
Ankara University ,\\
\ (oarslan@ankara.edu.tr)}
}
\renewcommand{\authorheading}
{Mehmet Niyazi \c{C}ankaya \ and \ Abdullah Yal\c{c}{\i}nkaya \ and \ \"{O}mer Alt{\i}nda\v{g} \ and \ Olcay Arslan } % --- on even running heads
%-------------------------------------------------------------------------------
% Just asking LaTeX to effectively do the title...
%-------------------------------------------------------------------------------
\maketitle
%-------------------------------------------------------------------------------
% Abstract of the article...
%-------------------------------------------------------------------------------
\begin{abstract}
The Burr III distribution is used in a wide variety of fields of lifetime data analysis, reliability theory, and financial literature, etc.
It is defined on the positive axis and has two shape parameters, say $c$ and $k$. These shape parameters make the distribution quite flexible.
They also control the tail behavior of the distribution. In this study, we extent the Burr III distribution to the real axis and also add
a skewness parameter, say $\varepsilon$, with epsilon-skew extension approach. When the parameters $c$ and $k$ have a relation such that
$ck \approx 1 $ or $ck < 1 $, it is skewed unimodal. Otherwise, it is skewed bimodal with the same level of peaks on the negative and positive sides of real line.
Thus, ESBIII distribution can capture fitting the various data sets even when the number of parameters are three. Location and scale form of this distribution are also given.
Some distributional properties of the new distribution are investigated. The maximum likelihood (ML) estimation method for the parameters
of ESBIII is considered. The robustness properties of ML estimators are studied and also tail behaviour of ESBIII distribution is examined.
The applications on real data are considered to illustrate the modeling capacity of this distribution
in the class of bimodal distributions.
\end{abstract}
%-------------------------------------------------------------------------------
% Keywords...
%-------------------------------------------------------------------------------
\begin{keywords}
Asymmetry; Burr III distribution; bimodality; epsilon skew; robustness; unimodality.
\end{keywords}
%-------------------------------------------------------------------------------
% AMS classifications...
%-------------------------------------------------------------------------------
\begin{ams}
49A05, 78B26.
\end{ams}
%-------------------------------------------------------------------------------
% Main text of the paper: sections, subsections and subsubsections,
% with theorems, lemmas, ..., figures, tables, ..., etc
%-------------------------------------------------------------------------------
\mainpaper % --- just a blank page and a \newpage command
%-------------------------------------------------------------------------------
% First section
%-------------------------------------------------------------------------------
\section{Introduction}
It is widely accepted that using flexible and heavy-tailed distributions is important to fit the data sets for practitioners in
application science. \cite{Burr42} proposed twelve different distributions which are quite flexible and useful to model the data sets.
Especially, Burr XII distribution is generally used as a flexible distribution \cite{Burr42,Burr68,Burr73,Rodriquez77,ali2015pak,Wingo83}.
However, Burr III distribution has a wider skewness and kurtosis region than that of Burr XII distribution \cite{Tadikamalla80}.
Burr III distribution has two shape parameters $c$ and $k$ which give the flexibility to the distribution.
Additionally, it has an ability that produces L-shaped or unimodal distribution when $ck \leq 1$ or $ck>1$
are satisfied, respectively \cite{Lindsay96}. Burr III distribution is defined on the positive axis,
so it is widely used in reliability studies \cite{wingo83,wang96,zimmer98,Zoraghietal12}, in the forest study
\cite{Goveetal08,Lindsay96} and in the hydrology study \cite{hao2009entropy}.
It is also used in the quality and control engineering, reliability analysis \cite{Shaoetal08,abd2012bayesian,azimi2013bayesian}.
%\begin{table}[htb]
%\footnotesize
%\caption{ML and MPS estimations when c=1,k=1}
%\label{tab:tablec1k1}
%\scalebox{0.95}{
%\begin{tabular}{l>{\hspace{0.7pc}}l>{\hspace{1pc}}l>{\hspace{1pc}}l>{\hspace{1pc}}l>{\hspace{1pc}}l>{\hspace{1pc}}l>{\hspace{1pc}}l>{\hspace{1pc}}l>{\hspace{1pc}}l>{\hspace{1pc}}l>{\hspace{1pc}}l>{\hspace{1pc}}l>{\hspace{1pc}}l}
%\hline
% & & \multicolumn{3}{c}{$\hat{\c}_{ML}$} & \multicolumn{3}{c}{$\hat{\k}_{ML}$} & \multicolumn{3}{c}{$\hat{\c}_{MPS}$} & \multicolumn{3}{c}{$\hat{\k}_{MPS}$} \\ \hline
%n & r & Mean & Bias & MSE & Mean & Bias & MSE & Mean & Bias & MSE & Mean & Bias & MSE\\ \hline
%20 & 14 & 0.9881& -0.0119& 0.5874& 1.0993& 0.0993& 0.0925& 1.1196& 0.1196& 2.7714& 0.9912& -0.0088& 0.0744 \\
% & 18 & 1.0682& 0.0682&0.0799& 1.0371& 0.0371& 0.0733& 0.9690& -0.0310& 0.0614& 1.0038& 0.0038& 0.0594 \\
%40 & 28 & 0.8668& -0.1332& 0.0554& 1.1088& 0.1088& 0.0461& 0.9993& -0.0007& 0.1258& 0.9928& -0.0072& 0.0352\\
% & 36 & 1.0161& 0.0161& 0.0278& 1.0255& 0.0255& 0.0336& 0.9760& -0.0240& 0.0263& 0.9980& -0.0020& 0.0301\\
%60 & 42 &0.8430& -0.1570& 0.0423& 1.1119& 0.1119& 0.0342& 0.9935& -0.0065& 0.0275& 0.9946& -0.0054& 0.0232\\
% & 54 &1.0011& 0.0011& 0.0169& 1.0227& 0.0227& 0.0218& 0.9814& -0.0186& 0.0168& 0.9978& -0.0022& 0.0201\\
%100& 70 &0.8241& -0.1759& 0.0404& 1.1137& 0.1137& 0.0253& 0.9907& -0.0093& 0.0154& 0.9963& -0.0037& 0.0137\\
% &90& 0.9879& -0.0121& 0.0096& 1.0192& 0.0192& 0.0127& 0.9856& -0.0144& 0.0098& 0.9971& -0.0029& 0.0119\\
%\hline
%\end{tabular}}
%\end{table}
%%Watkins (1999) Shao (2004)
In real life, most of the data sets are generally measured on the real axis and they may also have asymmetry as well.
There are well known distributions in the literature such as normal, student $t$, Laplace.
These are all symmetric distributions. Recently, many researchers have dealt with generalizing
the existing distributions to capture the data sets having asymmetry. There are various generalization
techniques which extent a symmetric distribution to asymmetric form. \cite{Azzalini85} proposed an
approach to construct skewed distribution and introduced firstly skew normal (SN) distribution.
Also \cite{AzzaliniCapi03} presented skew $t$ distribution. One can refer to \cite{Gentonbookskew04}
for Azzalini type skew distributions. \cite{Mud00} introduced so-called epsilon-skew method to
generalize the symmetric distributions to asymmetric form with a skewness parameter, say $\varepsilon$
such that $|\varepsilon|<1$. In that study, they presented the epsilon-skew normal (ESN) distribution.
If the skewness parameter $\varepsilon$ is chosen as zero, it reduces to the normal distribution.
From now on, this method has been applied to several well known distributions. \cite{Elsalloukhetal05}
considered the epsilon-skew extension of the exponential power (ESEP) distribution.
Special cases of the ESEP distribution reduce the ESN distribution \cite{Mud00} and epsilon-skew Laplace
(ESL) distribution \cite{Elsalloukh08}. The scale variant of ESEP distribution was considered
by \cite{Gomez07,ArslanGenc09}. \cite{Hassan13} took into account epsilon-skew extension of the reflected gamma distribution.
A comprehensive content about asymmetric distributions which include wide family of symmetric distributions as a special case can be found in \cite{ArelGom05}.
It should be noted that all the above distributions are unimodal skewed. However, one can encounter bimodal data sets in real life problems.
So, modeling the bimodality is another challenging problem. For this purpose, one can use mixed distributions.
However, there have been proposed bimodal distributions to model the bimodality directly. \cite{Hassan14,Andrade16,AndradeRat16,Cankayaetal15,Dexter15,Elal09,Donatella10,Genc13,Gui14,Has10,Hasel16,Rocha13,Cooray13,Jama11,Sham13,Rego12,Elal10,Arellanoetal10,Gomez11,Ahmed08,San08,Fam04,Eu02} proposed the distributions in the class of bimodality.
%There are some advantageous reasons for us to propose this distribution. In this context, the maximum likelihood estimators of parameters are obtained and their robustness property is investigated. As an indicator of robustness, the norm of influence function (IF) of ML estimators of parameters should be bounded. IF is defined as a linear transformation of score functions. To get IF, score function which is the derivative with respect to parameters of interest in the objective function obtained from the minus log-likelihood is used \cite{Hampel86}. We observe that the norm of IF of ML estimators for the location, scale, skewness and shape $k$ parameters are bounded when the shape parameter $c$ is assumed to be known. The opportunity in this distribution is that when just only one parameter is fixed, it is enough to approve having the robustness property of the ML estimators. So, the first reason is that this one is also an alternative distribution for the heavy-tailed candidates (such as epsilon skew Laplace, epsilon skew t distribution).\\
%
%
%There is a third opportunity of this distribution is that it can capture the bimodality when one wants to compare the generalized $t$ distribution in $\varepsilon$ skew form (ESGT) proposed by \cite{Gomez07,ArslanGenc09,Gomez12}. It is known that the special case of ESGT with $p=2$ and $q=\nu/2$ is the epsilon skew $t$ (EST) distribution. Since they are scale mixture of normal distributions, they are in class of heavy-tailed distributions. The univariate and the multivariate version of them in scale mixture and slash distribution are studied by \cite{Arslan09a,Arslan09b}. When reviewing the literature on bimodal and skew distributions, there are many different proposals. Some of these works are \cite{Cankayaetal15,Gui14,Genc13,Hassan14,Rocha13,Cooray13,Jama11,Sham13,Rego12,Elal10,Arellanoetal10,Gomez11,Ahmed08,San08,Fam04,Eu02}. \cite{Cankayaetal15} proposes the bimodal with different height. It is the bimodal extension of generalized gamma distribution (BEGG). \cite{Mud00} is an example for $\varepsilon-$ skew with non-heavy tailed distribution. In this study, we will propose same high bimodal and an asymmetric distribution at which $\varepsilon$ skew variant form is examined. Epsilon skew normal (ESN) in the class of thin-tailed distributions and epsilon skew Laplace (ESL) distribution in the class of heavy-tailed distributions are also used to make a comparison on modelling capacity between the proposed distribution and them in Section \ref{applications}. \\
%
%%As the fourth reason, it can be assumed that the error terms in linear or nonlinear regression models can have our distribution, since the ML estimators of parameters without parameter c of ESBIII distribution are robust to outlier in y-direction. Furthermore, the unimodality and bimodality are another advantage of this distribution for regression case, namely location, as we implied before.\\
%
%The last and major reason is that ESBIII has also skewness parameter, so it can handle the asymmetry in the data sets. With this property, it can be an alternative to skew distributions, such as skew normal distribution proposed by \cite{Azzalini85}. When the skewness parameter defined on $(-1,1)$ is greater than zero, the distribution will be skewed to positive axis, otherwise it will be skewed to negative axis. On the other hand, it is noted that when the skewness parameter is zero, the distribution is symmetric. So, it will be alternative to the symmetric heavy tailed distributions.\\
%In this paper, our aim is to propose a new flexible distribution on real axis and also modelling the symmetry and asymmetry of data set. For this objective, we extent the Burr III distribution to the real axis and also add a skewness parameter with the epsilon-skew extension approach. Anymore, we named it as the epsilon skew Burr III (ESBIII) distribution. Furthermore, we consider the location-scale transformation on ESBIII distribution to increase flexibility. So, the new distribution has five parameters which are location ($\mu$), scale ($\sigma$), skewness ($\epsilon$) and two shape ($c$, $k$) parameters.
In this paper, we first reflect the Burr III distribution to real axis and consider its epsilon-skew extension. Anymore, we call it as the epsilon-skew Burr III
(ESBIII) distribution. If the skewness parameter $\varepsilon \neq 0$, it can be unimodal skewed or bimodal skewed with same level of peaks, depending on the shape parameters.
The new distribution is not only unimodal, bimodal and skew but also candidate for efficient modelling when compared with ESEP distribution, because of the two
shape parameters in Burr III distribution. The scale version of ESEP is generalized $t$ distribution proposed by \cite{ArslanGenc09} and it has three parameters,
shape peakedness, tail and skewness parameters. In $ESBIII$, the number of parameters are three, however, it has the shape parameters $c$ and $k$
which help not only shape peakedness and tail behaviour but also the behaviour between peakedness and tail behaviour all together.
The order statistics distributions from ESBIII can be explicitly defined as an another advantage due to having an exact form of cumulative distribution function.
Since the skewed form of these two distributions are in same type, it is said that the the responsibility of parameter $\varepsilon$ are same for these two distributions.
The robustness property of this distribution is wide when compared with well known Student $t$ and its generalized skew version considered by \cite{ArslanGenc09}.
We also propose the location-scale transformation of ESBIII distribution.
Remainder of the paper is set up as follows. In Section 2, we present epsilon-skew extension of Burr III distribution. In Section 3,
we study some basic properties of the ESBIII distribution. In Section 4,
we obtain the maximum likelihood estimators (MLE) for the parameters of the ESBIII distribution.
Section $5$ investigates of the robustness property of MLE. In Section 6, some examples are given to show the performance of the ESBIII. The last section considers the conclusions.
\section{The Epsilon Skew Burr III Distribution}\label{B3definiton}
Consider the random variable $Z$ has a Burr III distribution with the density function
\begin{equation}
g(z)=ck z^{-(c+1)} (1+z^{-c})^{-(k+1)}.
\end{equation}
and the distribution function
\begin{equation}
G(z)=(1+z^{-c})^{-k}.
\end{equation}
where $z>0$ and the shape parameters $c>0$ and $k>0$ are real-valued parameters that identify the characteristic of the distribution such as dispersion,
peakedness, tail thickness. As noted earlier, when $ck \leq 1$ the Burr III distribution is L-shaped, when $ck>1$ the Burr III distribution is unimodal.\\
The following theorem states the epsilon skew extension of the Burr III distribution.
\begin{theorem}\label{B3theorem}
Let $Z$ be a continuous random variable distributed as a Burr III with the parameters $c$ and $k$. Let $U$ be
a discrete random variable with the following values with their probabilities,
\begin{equation}\label{Trv}
U=\left\{
\begin{array}{ll}
1+\varepsilon, & \frac{1+\varepsilon}{2} \\
-(1-\varepsilon), & \frac{1-\varepsilon}{2}
\end{array}
\right.
\end{equation}
\noindent where $\varepsilon \in (-1,1)$. Assume that $Z$ and $U$ are independent. Then, the distribution of the random variable
\begin{equation}\label{xvariable}
X=Z U
\end{equation}
\noindent will have the following distribution function
\begin{equation}\label{dfESB3}
F(x)=\left\{
\begin{array}{ll}
\frac{1-\varepsilon}{2}+(\frac{1+\varepsilon}{2})(1+(\frac{x}{1+\varepsilon})^{-c})^{-k} , & x \geq 0 \\
(\frac{1-\varepsilon}{2})(1-(1+(\frac{-x}{1-\varepsilon})^{-c})^{-k}) , & x<0
\end{array}
\right.
\end{equation}
and will have the following density function
\begin{equation}\label{pdfESB3}
f(x)=\frac{ck}{2} \bigg(\frac{sign(x)x}{1+sign(x)\varepsilon}\bigg)^{-(c+1)}\bigg(1+\bigg(\frac{sign(x)x}{1+sign(x)\varepsilon}\bigg)^{-c}\bigg)^{-(k+1)} , x \in \mathbf{R\backslash{0}}
\end{equation}
\noindent with the shape parameters $c>0$, $k>0$ and the skewness parameter $\varepsilon \in (-1,1)$.
\end{theorem}
\begin{proof}
%\smartqed
Using the distribution function technique, the CDF of $X$ for $x<0$ is
\begin{equation}\label{p1esb3}
F_{X}(x)=P(ZU\leq x)=\frac{1-\varepsilon}{2}\bigg\{1-F_{Z}\bigg(\frac{-x}{1-\varepsilon}\bigg)\bigg\}
\end{equation}
and for $x \geq 0$ is
\begin{equation}\label{p1esb3}
F_{X}(x)=P(ZU\leq x)=\frac{1-\varepsilon}{2}+\frac{1+\varepsilon}{2}\bigg\{F_{Z}\bigg(\frac{x}{1+\varepsilon}\bigg)\bigg\}.
\end{equation}
Taking the derivative of equation~\eqref{dfESB3} with respect to $x$ yields the density function of $X$ given in equation~\ref{pdfESB3}.
\qed
\end{proof}
\begin{definition}
The distribution of the random variable $X$ with the density function in equation \ref{pdfESB3} is called a Epsilon Skew Burr III ($ESBIII$) distribution.
\end{definition}
\begin{definition}
Suppose that $X \sim ESBIII(c,k,\varepsilon)$. Then, the random variable $Y=X\sigma + \mu$, $\sigma>0$, $\mu \in \mathbf{R}$ will have $ESBIII$ distribution
with the following density function ($Y \sim ESBIII(\mu,\sigma,c,k,\varepsilon)$)
\begin{equation}\label{locationscaleESBIII}
f(y)=\frac{ck}{2\sigma} \bigg(\frac{sign(\frac{y-\mu}{\sigma})\frac{y-\mu}{\sigma}}{1+sign(\frac{y-\mu}{\sigma})\varepsilon}\bigg)^{-(c+1)}\bigg(1+\bigg(\frac{sign(\frac{y-\mu}{\sigma})\frac{y-\mu}{\sigma}}{1+sign(\frac{y-\mu}{\sigma})\varepsilon}\bigg)^{-c} \bigg)^{-(k+1)} , y \in \mathbf{R\backslash{0}}.
\end{equation}
%\noindent where $\mu$ and $\sigma$ are the location and the scale parameters, respectively.
\end{definition}
The $ESBIII$ distribution has five parameters; $\mu$, $\sigma$, $\varepsilon$, $c$ and $k$. The parameters $\mu$ and $\sigma$
determine the location and the scale of the distribution, respectively. The parameter $\varepsilon$ controls the skewness
proportion on the left and right hand side of the location. The parameters $c$ and $k$ control the shape of the density together.
It is noted that when $\varepsilon=0$, it is symmetric unimodal or symmetric bimodal.\\
%Additionally, when $ck \leq 1$, the $ESBIII$ distribution is skewed unimodal; when $ck>1$, it is skewed bimodal.
\subsection{On the flexibility of ESBIII distribution}
Choosing special cases of the values of shape parameters $c$ and $k$ gives the flexible probability density functions (PDF).
For example, the shape of ESBIII distribution for some shape values can be similar to not only the epsilon skew Laplace distribution
but also epsilon skew t distribution. In this context, it is considered that the new distribution can model data sets which cannot
be precisely modelled by epsilon skew Laplace and epsilon skew t distributions, as exemplified by section \ref{applications}. The ESBIII has the properties of both leptokurtic
and platikurtic distributions. When $ck \approx 1$ or $ck < 1$, the PDF is unimodal. For $ck \approx 1$ or $ck < 1$ case,
when the difference between $c$ and $k$ is extremely large, the shape of distribution is platikurtic,
when the difference between $c$ and $k$ is extremely small, the shape of distribution is leptokurtic. The distributions
proposed by \cite{mineo2005software} and the references in \cite{Cankayaetal15} are in platikurtic-leptokurtic family as well.
When the values of $ck$ are close to $1$ from left hand side of $1$, the distribution has a peak being a needle. When $ck>1$,
the PDF is bimodal. The new distribution has also a skewness parameter. Thus, the base length of each group in the negative
and positive side of function becomes different. In Figure 1, some graphs are given to understand the behavior of the newly defined function for the different values of parameters.
The first and third columns in Figure $1$ represent the role of skewness parameter $\varepsilon$.
%%~\ref{fig:pdfexample}
\begin{figure}[!htb]
\advance\leftskip-1.8cm
\includegraphics[width=150mm, scale=.45]{pdfunimodal1}
\includegraphics[width=150mm, scale=.45]{pdfunimodal2}
\includegraphics[width=150mm, scale=.45]{pdfbimodal}
\caption{Examples of the PDF of the $ESBIII$ distribution for the different values of parameters, $\varepsilon=-0.5$ (left side), $\varepsilon=0$ (center) and $\varepsilon=0.5$ (right side).}
\label{fig:pdfexample}
\end{figure}
%The second opportunity to propose ESBIII distribution is that These properties of proposed distribution would be an advantage for the data analysis.\\
%%%Location and scale parameters, $\mu$ and $\sigma$, respectively is also For $ck=1$ case, even though the distribution is in the platikurtic class, The ESBIII distribution has a seat shape for the situation $ck=1$.
\begin{table}
\begin{center}
\caption{The skewness and the kurtosis values of $ESBIII$ distribution for the different values of parameters}
\label{tab:1}
\begin{tabular}{p{2.5cm} p{1.3cm} p{1.3cm} p{1.3cm} p{1.3cm} p{1.3cm}}
\hline
$c=20, k=0.20$ & $\varepsilon=0$ & $\varepsilon=0.2$ & $\varepsilon=0.4$ & $\varepsilon=0.6$ & $\varepsilon=0.8$ \\
\hline
Skewness & 0.0000 & 0.7453 & 1.0945 & 1.1553 & 1.1167 \\
Kurtosis & 1.1600 & 1.3021 & 1.4448 & 1.4071 & 1.2854 \\
\hline
\hline
$c=14, k=0.07$ & $\varepsilon=0$ & $\varepsilon=0.2$ & $\varepsilon=0.4$ & $\varepsilon=0.6$ & $\varepsilon=0.8$ \\
\hline
Skewness & 0.0000 & 0.9398 & 1.3801 & 1.4568 & 1.4082 \\
Kurtosis & 1.9481 & 2.1866 & 2.4262 & 2.3631 & 2.1587 \\
\hline
\hline
$c=10, k=0.10$ & $\varepsilon=0$ & $\varepsilon=0.2$ & $\varepsilon=0.4$ & $\varepsilon=0.6$ & $\varepsilon=0.8$ \\
\hline
Skewness & 0.0000 & 0.9591 & 1.4084 & 1.4866 & 1.4370 \\
Kurtosis & 2.0701 & 2.3236 & 2.5782 & 2.5111 & 2.2939 \\
\hline
\hline
$c=7, k=0.111$ & $\varepsilon=0$ & $\varepsilon=0.2$ & $\varepsilon=0.4$ & $\varepsilon=0.6$ & $\varepsilon=0.8$ \\
\hline
Skewness & 0.0000 & 1.0904 & 1.6013 & 1.6902 & 1.6338 \\
Kurtosis & 2.8707 & 3.2222 & 3.5752 & 3.4822 & 3.1809 \\
\hline
\hline
$c=5, k=0.20$ & $\varepsilon=0$ & $\varepsilon=0.2$ & $\varepsilon=0.4$ & $\varepsilon=0.6$ & $\varepsilon=0.8$ \\
\hline
Skewness & 0.0000 & 1.1760 & 1.7271 & 1.8230 & 1.7622 \\
Kurtosis & 4.2853 & 4.8100 & 5.3371 & 5.1982 & 4.7485 \\
\hline
\end{tabular}
\end{center}
\end{table}
In Table $1$, some values of the skewness and kurtosis are given for the different values of parameters. When the distribution
is the positive side skew, the sign of skewness parameter is positive. For negative side skewness, the skewness parameter is negative.
For both case, the kurtosis of distribution is positive. Since the skewness and kurtosis values are greater, it is obvious that the distribution is flexible.
\section{Some Properties of ESBIII Distribution}
The moments of $X \sim ESBIII(c,k,\varepsilon)$ are given by the following theorem.
\begin{theorem}\label{rthmoment}
Let $X \sim ESBIII(c,k,\varepsilon)$. The $rth$, $r \in \mathbf{R}$, noncentral moments are given by
\begin{equation}
\mathbf{E}(X^r)=\frac{k}{2} \mathbf{B}(1-\frac{r}{c},\frac{r}{c}+k) \{(1+\varepsilon)^{r+1}+(-1)^{-r}(1-\varepsilon)^{r+1}\}, c>r.
\end{equation}
\end{theorem}
\begin{proof}
%\smartqed
The $rth$ moment of $X$ can be easily obtained making the transformation $x=(1+\varepsilon)u^{-1/c}$ for negative part of integral
and the transformation $x=-(1-\varepsilon)u^{-1/c}$ for positive part of integral and using the following beta function
\begin{equation}\label{betafunc}
\int_{0}^{\infty}u^{-r/c}(1+u)^{-k-1}du=\mathbf{B}(1-\frac{r}{c},\frac{r}{c}+k),
\end{equation}
where $c>r$.
\qed
\end{proof}
\begin{corollary}
Let $X\sim ESBIII(c,k,\varepsilon)$. The expected value of $X$ is
\begin{equation}
\mathbf{E}(X)=\frac{k}{2} \mathbf{B}(1-\frac{1}{c},\frac{1}{c}+k) 4\varepsilon,~~ c>1.
\end{equation}
\end{corollary}
\begin{corollary}
Let $X\sim ESBIII(c,k,\varepsilon)$. The variance value of $X$ is
\begin{equation}
Var(X)=\frac{k}{2} \mathbf{B}(1-\frac{2}{c},\frac{2}{c}+k)(2+6\varepsilon^2)-k^2 \mathbf{B}^2(1-\frac{1}{c},\frac{1}{c}+k)4\varepsilon^2 ,~~ c>2.
\end{equation}
\end{corollary}
\begin{theorem}\label{CharaFunwithoutmusg}
Let $X \sim ESBIII(c,k,\varepsilon)$. The characteristic function is given by
\begin{equation}
\mathbf{E}(exp(itX))= \sum_{r=0}^{\infty} \frac{k (it)^r}{2r!} \mathbf{B}(1-\frac{r}{c},\frac{r}{c}+k) \{(1+\varepsilon)^{r+1}+(-1)^{-r}(1-\varepsilon)^{r+1}\}, c>r.
\end{equation}
\end{theorem}
\begin{proof}
The Taylor expansion of the function
$exp(itx) = \sum_{r=0}^{\infty} \frac{(itx)^r}{r!}$
can be used to calculate the integral $E(exp(itX))$.
\end{proof}
\begin{corollary}\label{CharaFunmusg}
Let $Y \sim ESBIII(\mu,\sigma,c,k,\varepsilon)$. The characteristic function (CF) for $\mu$ and $\sigma$ is given by
\begin{equation}
\mathbf{E}(exp(itY))= exp(it \mu) \sum_{r=0}^{\infty} \frac{k (it \sigma)^r}{2r!} \mathbf{B}(1-\frac{r}{c},\frac{r}{c}+k) \{(1+\varepsilon)^{r+1}+(-1)^{-r}(1-\varepsilon)^{r+1}\}.
\end{equation}
\end{corollary}
It is well known that $r!=\Gamma(r+1)$. Thus, the fractional form of CF will be proposed. It will be beneficial for the non integer order values for quantitative $r$, $r>0$.
Since the CDF is tractable, it is easy to propose the quantile function $x_q$ of $ESBIII$.
\begin{equation}\label{dfESB3}
F^{-1}(p)=x_q=\left\{
\begin{array}{ll}
(1+\varepsilon)\{ [\frac{2F(x)}{1+\varepsilon}-\frac{1-\varepsilon}{1+\varepsilon} ]^{-1/k} -1 \}^{-1/c} , & x \geq 0 \\
-(1-\varepsilon)[(1-\frac{2F(x)}{1-\varepsilon})^{-1/k}-1]^{-1/c} , & x<0.
\end{array}
\right.
\end{equation}
It is commonly used the R\'{e}nyi entropy for a random variable $X$ with PDF $f$. It is defined to be
\begin{equation}\label{Renyientropy}
\frac{1}{1-\alpha}\log\int_{-\infty}^{\infty}f^{\alpha}(x)dx.
\end{equation}
The following theorem is the R\'{e}nyi entropy for the ESBIII distribution.
\begin{theorem}
Let the random variable $X$ have an $ESBIII(0,1,c,k,\varepsilon)$ distribution. The R\'{e}nyi entropy of $X$ is given by
\begin{equation}\label{RenyientropyESBIII}
\frac{1}{1-\alpha}\log(I_1+I_2),
\end{equation}
where $\alpha>0$ and $\alpha \neq 1$,
\begin{equation}
I_1=(1-\varepsilon)(\frac{ck}{2})^{\alpha}\Gamma(\alpha(k+1)-\alpha(1+1/c)+1/c) \frac{\Gamma(\alpha(1+1/c)-1/c)}{c\Gamma(\alpha(k+1))}
\end{equation}
and
\begin{equation}
I_2=(1+\varepsilon)(\frac{ck}{2})^{\alpha}\Gamma(\alpha(k+1)-\alpha(1+1/c)+1/c) \frac{\Gamma(\alpha(1+1/c)-1/c)}{-c\Gamma(\alpha(k+1))}.
\end{equation}
\end{theorem}
\begin{proof}
The positive side of PDF $f$ with R\'{e}nyi entropy can be written as
$f_{+}^{\alpha}(x)=(\frac{ck}{2})^{\alpha}(\frac{x}{1+\varepsilon})^{-\alpha(c+1)}\bigg(1+(\frac{x}{1+\varepsilon})^{-c}\bigg)^{-\alpha(k+1)}$.
To get the result of the integral, $\frac{x}{1+\varepsilon}$ is taken to be $u$. After some straightforward calculation, the integral can be rewritten as following form,
\begin{eqnarray*}
% \nonumber to remove numbering (before each equation)
I_2&=& (1+\varepsilon)(\frac{ck}{2})^{\alpha}\int_{0}^{\infty} u^{-\alpha(c+1)}(1+u^{-c})^{-\alpha(k+1)}du \\ \nonumber
&=& (1+\varepsilon)(\frac{ck}{2})^{\alpha} \Gamma(\alpha(k+1)-\alpha(1+1/c)+1/c) \frac{\Gamma(\alpha(1+1/c)-1/c)}{-c\Gamma(\alpha(k+1))}.
\end{eqnarray*}
The same procedure can be done for the negative side of PDF $f$.
\qed
\end{proof}
\section{ML Estimators of Parameters in ESBIII and an Iterative Reweighting Approach for Computation}\label{MLELSESBIII}
Let $x_1,x_2,...,x_{n}$ be a random sample of size $n$ from a $ESBIII$ distributed population.
We would like to estimate the unknown parameters $\mu,\sigma,c,k,\varepsilon$. In this sense, we can prefer to use MLE and then the log-likelihood function is
\begin{eqnarray}
l &=& n \log(\frac{ck}{2\sigma})-(c+1)\sum_{i=1}^{n} \log(\frac{ sign(\frac{x_{i}-\mu}{\sigma}) (x_{i}-\mu)}{\sigma [1+sign(\frac{x_{i}-\mu}{\sigma}) \varepsilon]}) \\ \nonumber
&& -(k+1)\sum_{i=1}^{n} \log(1+(\frac{ sign(\frac{x_{i}-\mu}{\sigma}) (x_{i}-\mu)}{\sigma [1+sign(\frac{x_{i}-\mu}{\sigma}) \varepsilon]})^{-c}),\\ \nonumber
\end{eqnarray}
\noindent where $sign$ is the signum function and $sign(0)=1$.
The maximum likelihood estimators of the parameters $\mu,\sigma,c,k$ and $\varepsilon$ will be solution of the following equations
\begin{eqnarray}\label{nonlineareqmuu}
\frac{\partial l}{\partial \mu} &=& (c+1)\sum_{i=1}^{n} (x_{i}-\mu)^{-1}-c(k+1)\sum_{i=1}^{n} \frac{(x_{i}-\mu)^{-1}}{1+z_{i}^{c}}=0,
\end{eqnarray}
\begin{eqnarray}\label{nonlineareqsigma}
\frac{\partial l}{\partial \sigma} &=& \frac{nc}{\sigma}-c(k+1)\sum_{i=1}^{n} \frac{\sigma^{-1}}{1+z_{i}^{c}}=0,
\end{eqnarray}
\begin{eqnarray}\label{nonlineareqc}
\frac{\partial l}{\partial c} &=& \frac{n}{c}-\sum_{i=1}^{n} \log(z_{i}) +(k+1)\sum_{i=1}^{n} \frac{\log(z_{i})}{1+z_{i}^{c}}=0,
\end{eqnarray}
\begin{eqnarray}\label{nonlineareqk}
\frac{\partial l}{\partial k} &=& \frac{n}{k}-\sum_{i=1}^{n} \log\big(1+z_{i}^{-c}\big)=0,
\end{eqnarray}
\begin{eqnarray}\label{nonlineareqeps}
\frac{\partial l}{\partial \varepsilon} &=& (c+1)\sum_{i=1}^{n} (s_{i}+\varepsilon)^{-1}-c(k+1)\sum_{i=1}^{n} \frac{(s_{i}+\varepsilon)^{-1}}{1+z_{i}^{c}}=0,
\end{eqnarray}
\noindent where $s_{i}=sign(\frac{x_{i}-\mu}{\sigma}) $ and $z_{i}=\frac{s_{i}(x_{i}-\mu)}{\sigma [1+s_{i} \varepsilon]}$.\\
Since it is not possible to solve these equations analytically, the numerical methods should be used to obtain the $ML$ estimators.
Eqs. (\ref{nonlineareqmuu}) - (\ref{nonlineareqeps}) are solved in simultaneous time as an iteratively way, using the nonlinear solving
techniques in MATLAB 2013a. An example for solving an equation in Eqs. (\ref{nonlineareqmuu}) - (\ref{nonlineareqeps}) is as follow,
\begin{equation}\label{compustep}
\hat{\mu}^{(i+1)} := h(\mu,\hat{\sigma}^{(i)},\hat{c}^{(i)},\hat{k}^{(i)},\hat{\varepsilon}^{(i)}).
\end{equation}
As it is seen from equation \eqref{compustep}, the initial values of other parameters are provided.
The function $h$ representing the derivative of log-likelihood function with respect to parameter $\mu$ is solved
according to parameter $\mu$ via the nonlinear solving techniques in MATLAB 2013a. The same procedure was conducted for the other parameters of ESBIII.
While conducting procedure for solving, each step $i$ in equation \eqref{compustep} is computed until the prescribed value provided to finalize the
iterative algorithm is reached. The same procedure was conducted for the estimations of parameters of other distributions used to show
the fitting competence of ESBIII distribution. Kolmogorov-Smirnov statistic (KS) is a good tool to test whether or not the initial
values of parameters can be selected appropriately. The different initial values were tried until high value of KS is reached.
The approach can be more preferable when the computational issue for solving the nonlinear equations simultaneously has been arose.
\section{On The Robustness Properties of ESBIII Distribution}
\subsection{Robustness of ML estimators of parameters in ESBIII distribution}
%: Local and Global Robustness
%It is known that when the regularity conditions are satisfied, ML estimators are uniformly unbiased minimum variance (UMVUE), but the estimators must also have a robustness property.
ML estimators will be robust if the norm of influence function (IF) of ML estimators is bounded. IF is defined to be a linear transformation of the score vector.
As given in \cite{Hampel86}, IF for $\hat{\theta}$ is
\begin{equation}\label{IFdefinition}
IF(x;\hat{\theta})=-(E[\partial \psi(x;\boldsymbol \theta)/\partial \theta])^{-1}\psi(x;\boldsymbol \theta).
\end{equation}
\noindent It means that IF is proportional to the score functions, where $\theta = (\mu,\sigma,c,k,\varepsilon)$ is a vector of parameters.
$\psi$ is derivative of $-\log(f)$ with respect to the parameters and it is defined to be the score function.
It is noted that $-\log(f)$ can be regarded as an objective function $\rho$ in the context of robustness.
Here, the components of the local robustness of ML estimators are introduced.\\
\begin{definition}
Let $\psi$ be a score function derived by derivative of $-\log(f)$ with respect to parameters estimated. If $-\log(f)$ is differentiable,
then the function $\psi$ will be defined to be a score function of parameters \cite{Hampel86}.
\end{definition}
We will examine whether or not the score functions are bounded. When the limit of score function $\psi_{\theta}$ with respect to $x$ is bounded, ML estimators will be robust.\\
The functional forms of the score function corresponding to the parameters $\mu,\sigma,c,k,\varepsilon$ are
\begin{equation}\label{psimu}
\psi_{\mu}(x)=\frac{c+1}{x}-\frac{(k+1)c(\frac{sign(x)x}{1+sign(x)\varepsilon})^{-c}}{x(1+(\frac{sign(x)x}{1+sign(x)\varepsilon})^{-c})},
\end{equation}
\begin{equation}\label{psisigma}
\psi_{\sigma}(x)=c-\frac{(k+1)c(\frac{sign(x)x}{(1+sign(x)\varepsilon)})^{-c}}{1+(\frac{sign(x)x}{1+sign(x)\varepsilon})^{-c}},
\end{equation}
\begin{equation}\label{psic}
\psi_{c}(x)= \frac{1}{c}-\log(\frac{sign(x)x}{1+sign(x)\varepsilon})+\frac{(k+1)(\frac{sign(x)x}{1+sign(x)\varepsilon})^{-c}\log(\frac{sign(x)x}{1+sign(x)\varepsilon})}{1+(\frac{sign(x)x}{1+sign(x)\varepsilon})^{-c}},
\end{equation}
\begin{equation}\label{psik}
\psi_{k}(x)=\frac{1}{k}-\log\bigg(1+(\frac{sign(x)x}{1+sign(x)\varepsilon})^{-c} \bigg)
\end{equation}
\noindent and
\begin{equation}\label{psie}
\psi_{\varepsilon}(x)=\frac{sign(x)(c+1)}{1+sign(x)\varepsilon}-sign(x)\frac{(k+1)c(\frac{sign(x)x}{1+sign(x)\varepsilon})^{-c}}{(1+sign(x)\varepsilon)(1+(\frac{sign(x)x}{1+sign(x)\varepsilon})^{-c})}.
\end{equation}
$\underset{x \rightarrow \pm \infty}{lim} \psi_\mu(x) = 0$, $\underset{x \rightarrow \pm \infty}{lim} \psi_\sigma(x) = c$,
$\underset{x \rightarrow \pm \infty}{lim} \psi_c(x) = \infty$, $\underset{x \rightarrow \pm \infty}{lim} \psi_k(x) = 1/k$
and $\underset{x \rightarrow \pm \infty}{lim} \psi_\varepsilon(x) = \pm \frac{c+1}{1+sign(x)\varepsilon}$.
It is seen that the limit values of score functions for $\mu,k$ and $\varepsilon$ are finite,
but the limit value of $\psi_c(x)$ is infinite. Since the norm of influence function
$IF_{\mu,\sigma,k,\varepsilon}(x)$ of ML estimators for the parameters $\mu,\sigma,k$ and $\varepsilon$ is
bounded when $c$ is fixed and finite, the ML estimators for them are locally robust.
This shows us that the parameter $c$ must be fixed and finite. \cite{ArslanGenc09,Acitas13} and \cite{Lucas97} give the same discussion on the boundness of score function.
%If value $c$ is kept as fixed and finite, the score function $\psi_\sigma$ will be finite.
%\subsection{Robustness of ML estimators of parameters in ESBIII: Global Robustness}
\begin{definition}
Let $\psi$ be a score function. Suppose that there is a point $x_0$. For $0<x \leq x_0$, $\psi$ is (weakly) increasing.
For $x_0 < x < \infty$, $\psi$ is (weakly) decreasing. Then, $\psi$ is defined to be a redescending function \cite{Huber84}.
\end{definition}
\begin{enumerate}
\item $\rho(0)=0$ \cite{Huber84,ZhangLi98}.
%\item $x<0$ için $\rho(x)$ artmayan ve $x>0$ için $\rho(x)$ azalmayan bir fonksiyondur \cite{Huber84,ZhangLi98}.
\item $\underset{|x| \rightarrow \infty}{lim} \rho (x) = \infty$ \cite{Huber84,ZhangLi98}.
\item $\underset{|x| \rightarrow \infty}{lim} \frac{\rho (x)}{|x|} = 0$ \cite{Huber84}.
%\item $\psi(x) = \frac{d}{dx}\rho(x)$ $\mathbb{R}$'de süreklidir (Huber 1984).
\item For $0<x \leq x_0$, $\psi$ is weakly increasing. For $x_0 < x < \infty$, $\psi$ is weakly decreasing. There is a such a point $x_0$ \cite{Huber84}.
%\item $\rho(x)$ konvekstir. $\psi(x)=\frac{d}{dx}\rho(x)$ her yerde var ve $\psi(-\infty)<0<\psi(+\infty)$ (Zhang ve Li 1998).
\end{enumerate}
$\rho_{ESBIII}(x)=-\log(f)$. Let us show whether or not these conditions are satisfied.
\begin{enumerate}
\item $\rho_{ESBIII}(0)=0$ is not satisfied.
%\item $x<0$ için $\rho(x)$ artmayan ve $x>0$ için $\rho(x)$ azalmayan bir fonksiyondur. is satisfied, because ESBIII is on positive axis. It takes the different shapes for the distribution
\item $\underset{|x| \rightarrow \infty}{lim} \rho_{ESBIII} (x) = \infty$ is satisfied.
\item $\underset{|x| \rightarrow \infty}{lim} \frac{\rho_{ESBIII} (x)}{|x|} = 0$ is satisfied.
\item This condition is satisfied for only the MLE of parameter $\mu$.
\end{enumerate}
The condition $4$ is approved by the following lines.
It should be given some few comments. The derivatives of score functions with respect to $x$ for the parameters
$s,c,k,\varepsilon$ can not be solved according to $x$. Thus, it is not possible to find an exact
point $x_0$. From this result, it is said that the score functions of parameters $s,c,k,\varepsilon$
can not be in redescending function. Since a point $x_0$ can not be gotten,
it is not possible to imply that these functions are redescending.
\begin{proposition}
$\psi_\mu$ is a score function for parameter $\mu$. It can be possible to find a point $x_0$ at least for $\psi_\mu$.
$x_0={\frac {1}{a} \left( \,{\frac {-{c}^{2}k-{c}^{2}-ck+c+2+\sqrt {{c}^
{4}{k}^{2}+2\,{c}^{4}k+2\,{c}^{3}{k}^{2}+{c}^{4}+{c}^{2}{k}^{2}-2\,{c}
^{3}-2\,{c}^{2}k-3\,{c}^{2}}}{2(ck-1)}} \right) ^{-{c}^{-1}}}$,
\noindent where $a=\frac{sign(x)}{1+sign(x)\varepsilon}$.
\end{proposition}
Thus, it is possible to imply that the function $\psi_\mu$ is redescending.
\subsection{Tail behaviour of ESBIII distribution}
To call a heavy-tailed distribution or tail behaviour of a distribution, some definitions are proposed in \cite{Heavytail}:
%\begin{definition}\label{def1}
% Suppose that random variable $X$ has a CDF $F$ defined on $(0,\infty)$. If $E(exp(tX))=\infty$, for all $t$, then $F$ is a heavy-tailed.
%\end{definition}
%From equations 14 or 15, $E(exp(tX))=\infty$ is not satisfied for the ESBIII distribution. It is trivially hold.
\begin{definition}\label{def2}
Let $\bar{F}(x)$ be $1-F(x)$. If $\underset{x \rightarrow + \infty}{lim} exp(\lambda x)\bar{F}(x) = \infty$ for all $\lambda >0$, then $F(x)$ is a heavy-tailed.
\end{definition}
%\eqref{dfESB3}
From equation (5), the CDF is
$\frac{1-\varepsilon}{2}+(\frac{1+\varepsilon}{2})(1+(\frac{x}{1+\varepsilon})^{-c})^{-k}$ for $x \geq 0$. Here, the part $\frac{1-\varepsilon}{2}$ is negative side of PDF. This part is due to variable transformation. It leads to asymmetrization on a symmetric function. Thus, it can be hold on for the representation of negative side of PDF. From this clarification,
$(\frac{1+\varepsilon}{2})(1+(\frac{x}{1+\varepsilon})^{-c})^{-k}$
is a part for the positive side of CDF. In such a case,
$\bar{F}(x)=\frac{1+\varepsilon}{2}-(\frac{1+\varepsilon}{2})(1+(\frac{x}{1+\varepsilon})^{-c})^{-k}$.
Let us show the $\underset{x \rightarrow + \infty}{lim} exp(\lambda x)\bar{F}(x)$ for the ESBIII. As it is seen,
$\underset{x \rightarrow + \infty}{lim} exp(\lambda x)\bar{F}(x)=\infty$.
As a result, this property for heavy-tailedness is satisfied.
%\begin{definition}\label{def3}
%$\underset{x \rightarrow \pm \infty}{lim} \frac{\bar{F}(x+p)}{\bar{F}(x)}=1$, $p>0$.
%\end{definition}
%%\eqref{dfESB3}
%From equation (5), it is trivially hold.
Since $ESBIII$ distribution can be considered as a heavy-tailed distribution, the informative behaviour can be sensible degree when the data set includes one or more outliers.
In the following section, real data examples will be considered. Some useful comments can be given in the following lines.
As it is seen from Figure $1$, the platikurtic-leptokurtic property is hold for the ESBIII.
Thus, ESBIII can be a good candidate for the data modelling when there is a comparison test
between ESBIII and ESEP. When we look at the advantage of ESBIII, it is observed that ESBIII
can be preferable in the context of CDF performance, because the CDF of ESEP distributions
depends on the incomplete gamma function. For the CDF of ESBIII, we have an explicit analytical form,
which does make that KS test statistic of ESBIII is free from the computational error occurred by the
special function that is incomplete gamma function. The second derivative test is not examined for
the parameters of ESBIII distribution, because the analytical form of PDF is not easy to get the Hessian
matrix and its eigenvalues to check that the eigenvalues are negative for the parameters $c >0$ and $k>0$.
However, there is a good way to overcome this problem, because ESBIII has an explicit analytical form for
CDF, in other words, it is free from a special function such as incomplete gamma function. Thus,
this explicit analytical form will help us to control the efficient fitting performance of the ESBIII via KS test statistic.
%As it is seen from Figure $1$, the platikurtic-leptokurtic property is hold for the ESBIII. Thus, ESBIII can be a good candidate for the data modelling when there is a comparison test between ESBIII and ESEP. When we look at the advantage of ESBIII, it is observed that ESBIII can be preferable in the context of CDF performance, because the CDF of ESEP distributions depends on the incomplete gamma function. For the CDF of ESBIII, we have an explicit analytical form, which does make that KS test statistic of ESBIII is free from the computational error occurred by the special function that is incomplete gamma function.
%\underset{\xi \rightarrow 0}{lim}
%
%$\underset{x \rightarrow \pm \infty}{lim} \psi_\mu(x) = 0$
%
%$\lim_{x \rightarrow \pm \infty} \psi_\mu(x) = 0$
\section{Real Data Application}\label{applications}
In this section, two real data sets will be used to illustrate the modeling capability and robustness property of the proposed distribution.
%We used two data sets to test the fitting performance and robustness property of ESBIII distribution.
%A real data set is used and an outlier to real data set was added to test the fitting performance and robustness property of ESBIII distribution.
The bimodal distributions having the CDF are considered. Some of them are the $\varepsilon$-skew exponential distribution, its bimodal version with different height ($BEGG(\alpha,\beta,$ $\delta_0, \delta_1,\eta,\varepsilon,\mu,\sigma)$) considered by \cite{Elsalloukhetal05,Cankayaetal15}, respectively. The special case of the $\varepsilon$-skew exponential power distribution is the $\varepsilon$-skew Laplace distributions ($ESL$). The detailed discussion about the special cases of $BEGG$ distribution can be found in \cite{Cankayaetal15}. The generalized $t$ distribution in the sense of $\varepsilon-$skew form considered by \cite{ArslanGenc09} ($ESGT(p,q,\varepsilon,\mu,\sigma)$), the $\alpha$-skew Laplace distribution ($ASL(\alpha,\mu,\sigma)$) considered by \cite{Sham13} and the exponentiated sinh cauchy distribution ($ESC(\lambda,\beta,\mu,\sigma)$) considered by \cite{Cooray13} are used to model the data set.
%When the peakedness parameter $p$ is $2$, the $ESGT$ distribution is $\varepsilon$-skew ($EST$) distribution.
%%%Martin den oncesi
%The bimodal distributions having the CDF are considered. Some of them are the $\varepsilon$-skew exponential distribution, its bimodal version with different height ($BEGG(\alpha,\beta,\delta_0,\delta_1,\eta,\varepsilon,\mu,\sigma)$) considered by \cite{Elsalloukhetal05,Cankayaetal15}, respectively. The special cases of the $\varepsilon$-skew exponential power distribution are the $\varepsilon$-skew normal ($ESN$) and $\varepsilon$-skew Laplace distributions ($ESL$). The detailed discussion about the special cases of $BEGG$ distribution can be found in \cite{Cankayaetal15}. The generalized $t$ distribution considered by \cite{ArslanGenc09} ($ESGT(p,q,\varepsilon,\mu,\sigma)$) is used. When the peakedness parameter $p$ is $2$, the $ESGT$ distribution is $\varepsilon$-skew ($EST$) distribution. The $\alpha$-skew Laplace distribution ($ASL(\alpha,\mu,\sigma)$) considered by \cite{Sham13} is used. The exponentiated sinh cauchy distribution ($ESC(\lambda,\beta,\mu,\sigma)$) considered by \cite{Cooray13} is used.
%%%Martin den oncesi
%Iteratively reweighting algorithm widely accepted that this algorithm guarantees to convergence. \\
%Due to the fact that the analytical properties of function for ESGT and ESN and ESL distributions are appropriate to construct the iterative reweighing algorithm, we prefer to use the this algorithm.
\textbf{Example 1}: \cite{Arslan09a,Arslan09b} modelled the univariate cases of the data sets in cDNA microarray. They proposed parametric models. These models are used to fit the data sets. This data sets were also considered by \cite{Acitas13,PurdomHol05} for modelling data sets in the univariate case. In this study, we consider to model the cDNA microarray datasets. In these data sets, AT-matrix: 1375 genes $\times$ 118 drugs which are available at the web site
\begin{verbatim}http://discover.nci.nih.gov/
nature2000/data/selected_data/at_matrix.txt\end{verbatim}
are considered to test the performance of $ESBIII$.\\
In these data sets, the column one of gene-drug correlation data, which is named as "sid 487493, erythrocyte band 7
integral membrane protein [5,3:aa045112]", will be modelled by means of ESBIII distribution.
The candidates of $\varepsilon$-skew distribution family (ESEP) were proposed by \cite{Elsalloukhetal05} and \cite{Mud00}.
The scale mixture form of ESEP was considered by \cite{ArslanGenc09}, which can be considered as a $\varepsilon$-skew generalized $t$ distribution.
In this framework, we want to make a comparison among the distributions used in this work.
%the performance of $\varepsilon$-skew distribution family and ESBIII distribution.
%\begin{tiny}
%\begin{table}[htb]\label{esepesb3ex1}
%\begin{center}
%\caption{AY:Parameter estimates for the gene data set for the different parametric models}
%\begin{tabular}{c c c c c c c}
%\hline Parametric Models &$\hat{\mu}$ & $\hat{\sigma}$ & $\hat{\varepsilon}$ & $KS$ & $AIC$ \\ [0.5ex] % inserts table
%%heading
%\hline \hline % inserts single horizontal line
%ESBIII ($\hat{c}=2.0055, \hat{k}=0.7849$) & 0.0024 & 0.0770 & 0.0538 & 0.8959 & -403.5475 \\
%ESN ($\alpha=2$) & 0.0101 & 0.1133 & -0.0082 & 1.e-5 & -290.1165 \\
%ESL ($\alpha=1$) & 0.0115 & 0.0770 & -0.0009 & 0.0012 & -352.0729 \\
%ESGT($p=2,q=3/4$): EST & 0.0010 & 0.1063 & 0.0618 & 0.0074 & -149.6300\\
%BEGG($\hat{\alpha}=1,\hat{\beta}=1,\hat{\delta_0}=0.06,\hat{\delta_1}=0.35,\hat{\eta}=0.72$) & 0.0819 & 0.07 & 0.21 & 0.2062 & -69.7982 \\
%%ESC($\hat{\lambda}=1,\hat{\beta}=1) & & & & & \\
% \hline
%\end{tabular}
%\label{sheeptable}
%\end{center}
%\end{table}\end{tiny}
%\begin{tiny}
%\begin{table}[htb]\label{esepesb3outex1}
%\begin{center}
%\caption{AY:Parameter estimates for the gene data set with an outlier for the different parametric models}
%\begin{tabular}{c c c c c c c}
%\hline Parametric Models &$\hat{\mu}$ & $\hat{\sigma}$ & $\hat{\varepsilon}$ & $KS$ & $AIC$ \\ [0.5ex] % inserts table
%%heading
%\hline \hline % inserts single horizontal line
%ESBIII ($\hat{c}=1.9479, \hat{k}=0.7972$) & 0.0023 & 0.0780 & 0.0657 & 0.9422 & -383.5367 \\
%ESN ($\alpha=2$) & -0.0413 & 0.1362 & -0.3081 & 2.e-12 & -248.8255 \\
%ESL ($\alpha=1$) & 0.0120 & 0.0780 & -0.0411 & 2.e-4 & -351.9315 \\
%ESGT($p=2,q=3/4$): EST & -0.0039 & 0.1080 & 0.1026 & 0.0052 & -141.1141 \\
%BEGG($\alpha=1,\beta=1,\delta_0=0.06,\delta_1=0.35,\eta=0.72$) & 0.0819 & 0.07 & 0.21 & 0.2456 & -28.8669 \\
% \hline
%\end{tabular}
%\label{sheeptable}
%\end{center}
%\end{table}\end{tiny}
\begin{tiny}
\begin{table}[htb]\label{esepesb3ex1mnc}
\begin{center}
\caption{Parameter estimates for the microarray data set for the different parametric models}
\begin{tabular}{c c c c c c c}
\hline Parametric Models &$\hat{\mu}$ & $\hat{\sigma}$ & $\hat{\varepsilon}$ & $P(KS)$ & $AIC$ \\ [0.5ex] % inserts table
%heading
\hline \hline % inserts single horizontal line
ESBIII ($\hat{c}= 2.3826, \hat{k}= 0.7786$) & -0.0061 & 0.0770 & 0.0533 & 0.8478 & -432.4893 \\
ESL ($\alpha=1$) & 0.0130 & 0.0770 & 0.0484 & 0.0838 & -352.1553 \\
ESGT($p=1,q=3/4$) & 0.0440 & 0.0549 & -0.2773 & 0.0540 & -107.9998 \\
%ESGT($p=2,q=3/4$): EST & 0.0024 & 0.1058 & -0.0003 & 0.0011 & -162.7949 \\
%ESN ($\alpha=2$) & -0.0153 & 0.1025 & -0.1096 & 9.2e-26 & -313.7335 \\
\begin{tiny} BEGG($\hat{\alpha}=0.9824,\hat{\beta}=1,\hat{\delta_0}=0.0001,\hat{\delta_1}=0.4622,\hat{\eta}=0.7277$) \end{tiny} & 0.0105 & 0.0701 & 0.2301 & 0.2900 & -82.5059 \\
ASL($\hat{\lambda}=0.0781)$ & 0.0193 & 0.1022 &- & 0.5321 & -173.2460 \\
ESC($\hat{\lambda}=2.3591, \hat{\beta}=0.944) $ & 0.0051 & 0.1522 &- & 0.4037 & -158.4236 \\
\hline
\end{tabular}
\label{sheeptable}
\end{center}
\end{table}\end{tiny}
%\begin{figure}[!htb]
%\advance\leftskip-1cm
%\includegraphics[width=\textwidth, scale=.125]{PDFfigureEx1} %CDFexample
% \caption{Histogram of the data with the fitted pdf functions}
% \label{fig:cdfexample}
%\end{figure}
%\begin{figure}[!htb]
%\advance\leftskip-1cm
%\includegraphics[width=\textwidth, scale=.125]{CDFfigureEx1} %cdfexample
% \caption{Empirical CDF of the data with the fitted cdf functions}
% \label{fig:cdfexample}
%\end{figure}
\begin{figure}[!htb]
\centering
\begin{subfigure}{.55\linewidth}
\includegraphics[width=.85\textwidth]{PDFfigureEx1.eps}
\caption{Histogram of the data with the fitted PDF functions for microarray data}
\end{subfigure}%
\begin{subfigure}{.55\linewidth}
\includegraphics[width=.85\textwidth]{CDFfigureEx1.eps}
\caption{Empirical CDF of the data with the fitted CDF functions for microarray data}
\end{subfigure}
\end{figure}
\textbf{Example 2}: Martin Marietta data set is very popular in the literature. In this work, we will focus on fitting the explanatory
variable in the regression model. This variable is defined to be the excess rate of the Martin Marietta company.
\begin{tiny}
\begin{table}[htb]\label{esepesb3exMmnc}
\begin{center}
\caption{Parameter estimates for the Martin Marietta data set for the different parametric models}
\begin{tabular}{c c c c c c c}
\hline Parametric Models &$\hat{\mu}$ & $\hat{\sigma}$ & $\hat{\varepsilon}$ & $P(KS)$ & $AIC$ \\ [0.5ex] % inserts table
%heading
\hline \hline % inserts single horizontal line
ESBIII ($\hat{c}= 2.1241, \hat{k}= 0.7190$) & -0.0005 & 0.0652 & 0.0564 & 0.6136 & -242.7349 \\
ESL ($\alpha=1$) & -0.0050 & 0.0652 & -0.1434 & 0.1529 & -195.2337 \\
ESGT($p=1.2,q=1/4$) & -0.0096 & 0.0428 & 0.0699 & 0.4682 & -32.3817 \\
%ESGT($p=2,q=3/4$): EST & 0.0024 & 0.1058 & -0.0003 & 0.0011 & -162.7949 \\
%ESN ($\alpha=2$) & -0.0153 & 0.1025 & -0.1096 & 9.2e-26 & -313.7335 \\
\begin{tiny} BEGG($\hat{\alpha}=1.0824,\hat{\beta}=1.02,\hat{\delta_0}=0.2528,\hat{\delta_1}=0.0090,\hat{\eta}=0.7105$) \end{tiny} & 0.0201 & 0.0615 & 0.0301 & 0.1111 & -11.5659 \\
ASL($\hat{\lambda}=0.0001)$ & 0.0188 & 0.1018 &- & 0.5321 & -90.4340 \\
ESC($\hat{\lambda}=2.5921, \hat{\beta}=0.9860) $ & 0.0139 & 0.1534 &- & 0.4283 & -88.4240 \\
\hline
\end{tabular}
\label{sheeptable}
\end{center}
\end{table}\end{tiny}
\begin{figure}[!htb]
\centering
\begin{subfigure}{.55\linewidth}
\includegraphics[width=.85\textwidth]{PDFfigureExM.eps}
\caption{Histogram of the data with the fitted PDF functions for Martin Marietta data}
\end{subfigure}%
\begin{subfigure}{.55\linewidth}
\includegraphics[width=.85\textwidth]{CDFfigureExM.eps}
\caption{Empirical CDF of the data with the fitted CDF functions for Martin Marietta data}
\end{subfigure}
\end{figure}
%\begin{figure}[!htb]
%\centering
% \begin{subfigure}{.55\linewidth}
% \includegraphics[width=.85\textwidth]{PDFfigureExMden.eps}
% \caption{Histogram of the data with the fitted pdf functions for Martin Marietta data YUKSEKLIK}
% \end{subfigure}%
%% \begin{subfigure}{.55\linewidth}
%% \includegraphics[width=.85\textwidth]{CDFfigureExM.eps}
%% \caption{Empirical cdf of the data with the fitted cdf functions for Martin Marietta data}
%% \end{subfigure}
%\end{figure}
%%%outlier%%%outlier%%%outlier%%%outlier%%%outlier%%%outlier
%\begin{tiny}
%\begin{table}[htb]\label{esepesb3ex1mnc}
%\begin{center}
%\caption{Parameter estimates for the gene data set with outlier for the different parametric models}
%\begin{tabular}{c c c c c c c}
%\hline Parametric Models &$\hat{\mu}$ & $\hat{\sigma}$ & $\hat{\varepsilon}$ & $KS$ & $AIC$ \\ [0.5ex] % inserts table
%%heading
%\hline \hline % inserts single horizontal line
%ESBIII ($\hat{c}= 2.2882, \hat{k}= 0.8053$) & -0.0062 & 0.0770 & 0.0645 & 0.8931 & -402.6823 \\
%ESL ($\alpha=1$) & 0.0130 & 0.0770 & -0.0083 & 0.0341 & -355.0987 \\
%ESGT($p=2,q=3/4$): EST & -0.0446 & 0.1075 & 0.3095 & 1.7027e-05 & -156.1102 \\
%ESN ($\alpha=2$) & -0.0387 & 0.1189 & -0.2829 & 1.0542e-32 & -281.0359 \\
%\begin{tiny} BEGG($\hat{\alpha}=0.9824,\hat{\beta}=1,\hat{\delta_0}=0.0002,\hat{\delta_1}=0.4622,\hat{\eta}=0.7277$) \end{tiny} & 0.0805 & 0.0701 & 0.2201 & 0.2999 & -50.7945 \\
%ASL($\hat{\lambda}=0.0795)$ & 0.0199 & 0.1012 &- & 0.5549 & -160.3194 \\
%ESC($\hat{\lambda}=2.3591, \hat{\beta}=0.914) $ & 0.0169 & 0.1522 &- & 0.4074 & -148.4835 \\
% \hline
%\end{tabular}
%\label{sheeptable}
%\end{center}
%\end{table}\end{tiny}
%\begin{figure}[!htb]
%\advance\leftskip-1cm
%\includegraphics[width=\textwidth, scale=.125]{CDFfigureEx1outlier} %cdfexample
% \caption{Empirical cdf of the data with the fitted cdf functions with one outlier}
% \label{fig:cdfexample}
%\end{figure}
%%%outlier%%%outlier%%%outlier%%%outlier%%%outlier
%\begin{figure}[H]
%\centering
% \begin{subfigure}%{.7\linewidth}
% \includegraphics[width=.99\textwidth]{pdfexample.eps}
% \caption{Histogram of the data with the fitted densities}
% \end{subfigure}%
% \begin{subfigure}%{.7\linewidth}
% \includegraphics[width=.99\textwidth]{cdfexample.eps}
% \caption{Empirical cdf of the data with the fitted cdf functions}
% \end{subfigure}%
% %\caption{Measurement functions with $q$ parameters}
%\end{figure}
%
%\begin{figure}[H]
%\centering
% \begin{subfigure}{.7\linewidth}
% \includegraphics[width=.99\textwidth]{pdfexampleoutlier.eps}
% \caption{Histogram of the data with the fitted densities}
% \end{subfigure}%
% \begin{subfigure}{.7\linewidth}
% \includegraphics[width=.99\textwidth]{cdfexampleoutlier.eps}
% \caption{Empirical cdf of the data with the fitted cdf functions}
% \end{subfigure}%
% %\caption{Measurement functions with $q$ parameters}
%\end{figure}
%%%\ref{esepesb3ex1} \ref{esepesb3outex1}
Tables 2 and 3 show that ESBIII distribution fits better than the $\varepsilon$-skew candidate ones, such as ESL, ESGT, BEGG, ASL and ESC distributions.
It is considered that the parameters $c$ and $k$ produce more flexibility when compared with ESL, ESGT, BEGG, ASL and ESC especially.
The parameters $c$ and $k$ of ESBIII produce the flexible behaviour for the peakedness and tail thickness both together.
In addition to this property, the interspace between peakedness and tail of ESBIII can also be flexible,
because we have two shape parameters $c$ and $k$. This property can be considered as an advantage for modelling capacity of the distribution. \\
Tables 2 and 3 show the estimated values of parameters in the used models, $P(KS)$ that is the probability value (p-value) of KS statistic and
Akaike information criterion (AIC) values. It is known that the more p-value of KS statistic is highest, the more modelling capacity is best.
When the parametric models are ordered according to the modelling capacity from highest to lowest one,
ESBIII, ASL, ESC, BEGG, ESL and ESGT distributions are given in order for microarray data.
ESBIII distribution has the high p-value of KS test. Why it can model the data very efficiently is that
the function can capture the frequency as a flexible way. It can be considered that the parameters $c$ and $k$ keep their property that is shapeness.
When the parametric models are ordered according to the modelling capacity from highest to lowest one,
ESBIII, ASL, ESGT, ESC, ESL and BEGG distributions are given in order for Martin Marietta data.
Since ESBIII is in the class of the heavy-tailed distributions, it can model the data in tail.
The bimodality property of ESBIII distribution plays an important role in fitting the data set as well.
Same interpretation on modelling capacity can not be said for the other distributions.
As it is seen, the bimodality and heavy-tailedness are advantage for fitting.
%, because the tail behaviour of them is not capable as that of ESBIII does according the numerical results. The behaviour of the tail is an In other side, the occurred frequencies under function can be fitted and then the information gains. indicator for the modelling capacity of a distribution, which shows why the p-value of KS for ESN is extremely lowest when compared with the ESL, EST and BEGG distributions.
%%%martin oncesi%%%martin oncesi%%%martin oncesi%%%martin oncesi
%Tables 2 and 3 show that ESBIII distribution fits better than the $\varepsilon$-skew candidate ones, such as ESN, ESL, EST and BEGG. It is thought that the parameters $c$ and $k$ produce more flexibility when compared with ESN, ESL, EST and BEGG especially. The parameters $c$ and $k$ of ESBIII produce the flexible behaviour for the peakedness and tail thickness both together. In addition to this property, the interspace between peakedness and tail of ESBIII can also be flexible, because we have two shape parameters $c$ and $k$. This property is an advantage for modelling capacity of the distribution. Tables 2 and 3 show the estimated values of parameters in the used models, Kolmogorov-Smirnov statistic (KS) and Akaike information criterion (AIC) values. The more p-value of KS statistic is highest, the more modelling capacity is best. When the parametric models are ordered according to the modelling capacity from highest to lowest one, ESBIII, BEGG, EST, ESL and ESN distributions are given in order. ESBIII distribution has the high p-value of KS even if an outlier is added to data set. Why it can model the data with an outlier is that the function can capture the frequency wherever data is on axis. In other words, ESBIII is in the class of the heavy-tailed distributions. So, it can model the data in tail. In other side, the occurred frequencies under function can be fitted and then the information gains. Same interpretation on modelling capacity can not be said for the ESN, ESL and EST distributions, because the tail behaviour of them is not capable as that of ESBIII does. The behaviour of the tail is an indicator for the modelling capacity of a distribution, which shows why the p-value of KS for ESN is extremely lowest when compared with the ESL, EST and BEGG distributions. The bimodality property of ESBIII distribution plays an important role on fitting the data set as well.
%%%martin oncesi%%%martin oncesi%%%martin oncesi%%%martin oncesi
%%%\bibitem{NadKot06} Nadarajah S. Kotz S. (2006). q exponential is a Burr distribution. Physics Letters A 359 (2006) 577-579
%\begin{tiny}
%\begin{table}[htb]\label{esepesb3ex2}
%\begin{center}
%\caption{Parameter estimates for the gene data set for the different parametric models}
%\begin{tabular}{c c c c c c c}
%\hline Parametric Models &$\hat{\mu}$ & $\hat{\sigma}$ & $\hat{\varepsilon}$ & $KS$ & $AIC$ \\ [0.5ex] % inserts table
%%heading
%\hline \hline % inserts single horizontal line
%ESBIII ($\hat{c}=., \hat{k}=0.$) & 0. & 0. & 0. & 0. & -. \\
%ESN ($\alpha=2$) & 0. & 0. & -0. & . & -. \\
%ESL ($\alpha=1$) & 0. & 0. & -0. & 0. & -. \\
%ESGT($p=2,q=$):EST & 0. & 0. & 0. & 0. & -.\\
%BEGG($\alpha=,\beta=,\delta_0=0.,\delta_1=0.,\eta=0.$) & 0. & 0. & 0. & 0. & -. \\
% \hline
%\end{tabular}
%\label{sheeptable}
%\end{center}
%\end{table}\end{tiny}
%
%
%\begin{tiny}
%\begin{table}[htb]\label{esepesb3outex2}
%\begin{center}
%\caption{Parameter estimates for the gene data set with an outlier for the different parametric models}
%\begin{tabular}{c c c c c c c}
%\hline Parametric Models &$\hat{\mu}$ & $\hat{\sigma}$ & $\hat{\varepsilon}$ & $KS$ & $AIC$ \\ [0.5ex] % inserts table
%%heading
%\hline \hline % inserts single horizontal line
%ESBIII ($\hat{c}=., \hat{k}=0.$) & 0. & 0. & 0. & 0. & -. \\
%ESN ($\alpha=2$) & 0. & 0. & -0. & . & -. \\
%ESL ($\alpha=1$) & 0. & 0. & -0. & 0. & -. \\
%ESGT($p=2,q=$):EST & 0. & 0. & 0. & 0. & -.\\
%BEGG($\alpha=,\beta=,\delta_0=0.,\delta_1=0.,\eta=0.$) & 0. & 0. & 0. & 0. & -. \\
% \hline
%\end{tabular}
%\label{sheeptable}
%\end{center}
%\end{table}\end{tiny}
\cite{Lucas97} and \cite{ArslanGenc09} prefer to use the fixed values of parameters $p$ and $q$ as a tuning constant in robustness,
because the ML estimators of the parameters are not robust owing to fact that the IF of them is unbounded. For our case, since the
score function for the parameter $c$ is infinite, IF of ML estimators of parameters will be unbounded. We prefer to estimate it,
because it is clearly seen that the p-value of KS statistic shows the competence of our flexible distribution.
Using both the robustness of ML estimators for the parameters of distribution and the goodness of fit test representing the
modelling capacity of proposed distribution together is suggested. For our real data examples,
examining whether or not the p-value of KS statistic is high or the value of AIC is small will be a good indicator to
approve the modelling capacity of proposed distribution. Since the goodness of fit tests (KS and AIC) gives the satisfied
results, the parameter $c$ was also estimated. In fact, IF of ML estimators for the parameters
$\mu$, $\sigma$, $c$, $k$ and $\varepsilon$ is unbounded, however there is another point it should be taken in account,
which is that the modelling capacity of the proposed model is very high when the p-values of KS statistic are considered.
If the parameter $c$ was considered to be fix, the estimated value for it would be wonderful choice, because ESBIII has a
high p-value of KS statistic, which means that the high KS value or small AIC value show the best fitting performance on data set.
The robustness property of ESBIII also supports the performance of ESBIII.
%%%martinden once%%%martinden once%%%martinden once%%%martinden once
%\cite{Lucas97} and \cite{ArslanGenc09} prefer to use the fixed values of parameters $p$ and $q$ as a tuning constant in robustness, because the ML estimators of the parameters are not robust owing to fact that the IF of them is unbounded. For our case, since the score function for the parameter $c$ is infinite, IF of ML estimators of parameters will be unbounded. We prefer to estimate it, because it is clearly seen that the p-value of Kolmogorov-Smirnov statistic shows the competence of our flexible distribution. It can be suggested using robustness of ML estimators for the parameters of distribution and the goodness of fit test representing the modelling capacity of proposed distribution both together. For our real data examples with an outlier, examining whether or not the p-value of Kolmogorov-Smirnov statistic is high will be a good indicator to approve the modelling capacity of proposed distribution. Since the goodness of fit test gives the satisfied results, the parameter $c$ was also estimated. In fact, IF of ML estimators for the parameters $\mu$, $\sigma$, $c$, $k$ and $\varepsilon$ is unbounded, however there is another point it should be taken in account, which is that the modelling capacity of the proposed model is very high when the p-values of KS statistic are considered. If the parameter $c$ was considered to be fix, the estimated value for it would be wonderful choice, because ESBIII has a high p-value of KS statistic, which means that the high KS value means the best fitting performance on data set.
%%%martinden once%%%martinden once%%%martinden once
\section{Conclusions}\label{conclusions}
In this paper, an extension to real axis for the Burr III distribution has been proposed. The main goal at supplying the new distribution is that Burr III distribution has heavy tailed property. An extension to real axis already preserves the heavy tailed (robustness) property of this distribution. A skewness parameter was also added to the extended Burr III. The skewness parameter provides another flexibility when the positive side and negative side of real axis have different frequency. In this context, ESBIII distribution is one in the class of the heavy-tailed candidates and has a skewness parameter as well.
The parameters $c$ and $k$ in Burr III distribution are the shape parameters, which gives the flexibility on function. It is readily seen that some values of them produce unimodal, bimodal, leptokurtic and platikurtic distributions. The skewness and the kurtosis values of ESBIII distribution for the different values of parameters also approve the competency on flexibility of the proposed distribution. As a result, the ESBIII copes with skewness, heavy-tailedness, bimodality by means of just only three parameters.
%It is known that ML estimators for any distribution are asymptotically normal distributed. We have provided the asymptotic variance-covariance matrix of ML estimators for the proposed distribution. Since the extended Burr III distribution does also have skewness parameter, we have another advantage to fit the data which are unequal frequency in real axis, as implied previously.
The some properties of ESBIII were introduced; such as the R\'{e}nyi entropy and the $rth$ moment of X and the characteristic function were calculated.
The ML estimators of the parameters in ESBIII have been obtained.
The local robustness property of ML estimators for the parameters in ESBIII was examined and also ESBIII was examined to see the heavy-tailedness property.
Since we have the ML estimators for the parameters in ESBIII, the real data examples can be applied to see the performance of the distribution.
For this purpose, two real data examples were given. It was observed that the p-values of KS statistic for ESBIII were very high for the considered data sets.
We can assert that having shape parameters $c$ and $k$ which can work both together to construct the peakedness and the tail behaviour of the distribution produces
a flexible distribution. The skewness parameter $\varepsilon$ helps to model the uneqaul frequency in positive and negative part of real axis.
In other words, adding the skewness parameter $\varepsilon$ helps to generate a model which can capture the skewness in data.
For another study, one can make a comparison between the flexibility performance of Azzalini type skewness of the extended Burr III
distribution and $\varepsilon$ skew version proposed in this study. In ongoing paper, Shannon, Tsallis, R\'{e}nyi and hybrid entropies of distributions
\cite{JizbaKorbel16,CankayaKorbel17} will be used as a tool to test how much the degree of heavy-tailedness can be produced by some distributions.
%%%%yapýlcaklar
%example 2 yapýlcak
%pdf cizimlerinde legend koncak belkide pdfler kaldýrýlmalý
%\begin{figure}[H]
%\centering
% \begin{subfigure}{.7\linewidth}
% \includegraphics[width=.99\textwidth]{pdfexample.eps}
% \caption{Histogram of the data with the fitted densities}
% \end{subfigure}%
% \begin{subfigure}{.7\linewidth}
% \includegraphics[width=.99\textwidth]{cdfexample.eps}
% \caption{Empirical cdf of the data with the fitted cdf functions}
% \end{subfigure}%
% %\caption{Measurement functions with $q$ parameters}
%\end{figure}
%
%\begin{figure}[H]
%\centering
% \begin{subfigure}{.7\linewidth}
% \includegraphics[width=.99\textwidth]{pdfexampleoutlier.eps}
% \caption{Histogram of the data with the fitted densities}
% \end{subfigure}%
% \begin{subfigure}{.7\linewidth}
% \includegraphics[width=.99\textwidth]{cdfexampleoutlier.eps}
% \caption{Empirical cdf of the data with the fitted cdf functions}
% \end{subfigure}%
% %\caption{Measurement functions with $q$ parameters}
%\end{figure}
%\begin{center}
%\begin{figure}[!]
%%\advance\leftskip-1cm
%\includegraphics[width=\textwidth, scale=.15]{pdfexample}
% \caption{Histogram of the data with the fitted densities}
% \label{fig:pdfexample}
%\end{figure}
%\end{center}
%\begin{figure}[!htb]
%\advance\leftskip-1cm
%\includegraphics[width=\textwidth, scale=.15]{pdfexampleoutlier}
% \caption{Histogram of the data with an outlier together with the fitted densities}
% \label{fig:pdfexampleoutlier}
%\end{figure}
%\begin{figure}[!htb]
%\advance\leftskip-1cm
%\includegraphics[width=\textwidth, scale=.15]{cdfexampleoutlier}
% \caption{Empirical cdf of the data with an outlier together with the fitted cdf functions}
% \label{fig:cdfexampleoutlier}
%\end{figure}
%-------------------------------------------------------------------------------
% Acknowledgments...
%-------------------------------------------------------------------------------
%\begin{acknowledgments}
%This work has been supported by the grant number xyz from ABCD.\break
%We also acknowledge the valuable suggestions from Prof.\ EFG and the referees.
%\end{acknowledgments}
%-------------------------------------------------------------------------------
% References...
%-------------------------------------------------------------------------------
\begin{thebibliography}{8}
%\input{referenc}
\bibitem{abd2012bayesian} Abd-Elfattah, A.M. and Alharbey, A.H., Bayesian estimation for Burr distribution type III based on trimmed samples, ISRN Applied Mathematics, 2012
\bibitem{Hassan14} Abdulah E., Elsalloukh H., Bimodal Class based on the Inverted Symmetrized Gamma Distribution with Applications. J. Stat. Appl. Pro. 3, No. 1, 1-7, 2014
\bibitem{Andrade16} Andrade de B.B., Bolfarine H., Siroky A.N., Random number generation and estimation with the bimodal asymmetric power-normal distribution, Journal of
Statistical Computation and Simulation, 86:3, 460-476, 2016
\bibitem{Hassan13} Abdulah E. and Elsalloukh H., Analyzing skewed data with the epsilon skew gamma distribution, Journal of Statistics Applications and Probability, 2(3), 195-202, 2013.
\bibitem{Acitas13} Ac{\i}ta\c{s} S., Kasap, P. \c{S}eno\v{g}lu, B., Arslan O., One-step M-estimators: Jones and Faddys skewed t-distribution, Journal of Applied Statistics, 1-15, 2013
\bibitem{Ahmed08} Ahmed S.E., Goria M.N., Hussein A. Gamma mixture: bimodality, inflexions and L-moment, Communications in Statistics - Theory and Methods. 37(8):1147-1161, 2008
\bibitem{ali2015pak} Ali A., Hasnain S.A., Ahmad, M., Modified Burr III Distribution Properties and Applications, Pak. J. Statist., 31(6) 697-708, 2015
\bibitem{AndradeRat16} Andrade B.B., Rathie P.N., Fitting asymmetric bimodal data with selected distributions, Journal of Statistical Computation and Simulation 86:16,
3205-3224, 2016
\bibitem{ArelGom05} Arellano-Valle, R. B., Gomez, H. W., Quintana, F. A., Statistical inference for a general class of asymmetric distributions, J. Statist. Plan. Infer. 128:427443, 2005
\bibitem{Azzalini85} Azzalini A., A class of distributions which includes the normal ones, Scandinavian journal of statistics. (12)2 171-178, 1985
\bibitem{AzzaliniCapi03} Azzalini A., Capitanio A., Distributions generated by perturbation of symmetry with emphasis on a multivariate skew t distribution, J. Roy. Statist. Soc. Ser. B, 65, 367389, 2003
\bibitem{Arellanoetal10} Arellano-Valle R. B., Cort{\'e}s M. A., G{\'o}mez H{\'e}ctor W. An extension of the epsilon-skew-normal distribution, Communications in Statistics - Theory and Methods, 39(3),912-922, 2010
\bibitem{Arslan09a} Arslan O., Maximum likelihood parameter estimation for the multivariate skew slash distribution, Statistics \& Probability Letters, 79(20), 2158-2165, 2009a
\bibitem{Arslan09b} Arslan O., An alternative multivariate skew Laplace distribution: Properties and estimation, Statistical papers, 51:865-887, 2009b
\bibitem{ArslanGenc09} Arslan O., Gen\c{c} A.Ý. The skew generalized t distribution as the scale mixture of a skew exponential power distribution and its applications in robust estimation, Statistics, 43(5), 481-498, 2009
\bibitem{azimi2013bayesian} Azimi R., Yaghmaei F., Bayesian estimation for the burr type III distribution under type II Doubly censored data, International Journal of Advanced Statistics and Probability, 1(1) 1-3, 2013
\bibitem{Burr42} Burr I.W., Cumulative frequency functions, The Annals of mathematical statistics, 13(2), 215-232, 1942
\bibitem{Burr68} Burr I.W., Cislak P.J., On a general system of distributions I. Its curve-shape characteristics II. The sample median, Journal of the American Statistical Association, Vol 63. No. 322, 627-635, 1968
\bibitem{Burr73} Burr I.W., Pameters for a general system of distributions to match a grid of $\alpha$ and $\alpha$, Communications in Statistics-Theory and Methods Vol 2. No. 1. 1-21, 1973
\bibitem{Cooray13} Cooray K., Exponentiated Sinh Cauchy Distribution with Applications, Communications in Statistics - Theory and Methods, 42(21), 3838-3852, 2013
\bibitem{Cankayaetal15} \c{C}ankaya M.N., Bulut Y.M., Do\u{g}ru F.Z., Arslan O., A Bimodal Extension of the Generalized Gamma Distribution, Revista Colombiana de Estad{\'\i}stica, 38(2), 353--370, 2015
\bibitem{CankayaKorbel17} \c{C}ankaya M.N., Korbel J., On statistical properties of JizbaArimitsu hybrid entropy, Physica A: Statistical Mechanics and its
Applications, 2017. http://dx.doi.org/10.1016/j.physa.2017.02.009. (In Press).
\bibitem{Dexter15} Dexter O. C., Some Skew-Symmetric Distributions Which Include the Bimodal Ones, Communications in Statistics - Theory and Methods, 44:3, 554-563, 2015
\bibitem{Donatella10} Donatella V., van Dorp J.R., On a bounded bimodal two-sided distribution fitted to the Old-Faithful geyser data, Journal of Applied Statistics, 40:9, 1965-1978, 2013.
\bibitem{Elsalloukh08} Elsalloukh H., The epsilon-skew Laplace distribution, The Proceedings of the American Statistical Association, Biometrics Section, Denever Colorado, 2008.
\bibitem{Elal10} Elal-Olivero D., Alpha-skew-normal distribution, Proyecciones(Antofagasta), 29(3), 224-240, 2010
\bibitem{Elal09} Elal-Oliveroa D., Gómez H.W., Fernando A., Quintanac Bayesian modeling using a class of bimodal skew-elliptical distributions, Journal of Statistical Planning and Inference, 139(4), 2009.
\bibitem{Heavytail} Embrechts P., Kluppelberg C., Mikosch T., Modelling Extremal Events for Insurance and Finance, Springer, Heidelberg, 1997.
\bibitem{Eu02} Eugene N., Lee C., Famoye, F. Beta-normal distribution and its applications, Communications in Statistics - Theory and methods, 31(4), 497-512, 2002
\bibitem{Elsalloukhetal05} Elsalloukh H., Guardiola J. H., Young M., The epsilon-skew exponential power distribution family, Far East Journal of Theoretical Statistics, 17(1) 97-107, 2005
\bibitem{Fam04} Famoye F., Lee C., Eugene N., Beta-normal distribution: Bimodality properties and application, Journal of Modern Applied Statistical Methods, 3(1), 85-103, 2004
\bibitem{Genc13} Genc A. I., A skew extension of the slash distribution via beta-normal distribution, Statistical Papers, 54(2), 427--442, 2013
\bibitem{Gentonbookskew04} Genton M.G., Skew-Elliptical Distributions and Their Applications: A Journey Beyond Normality, Chapman \& Hall/CRC, 2004.
\bibitem{Gomez07} Gomez H.W., Torres F.J., Bolfarine H., Large-sample inference for the epsilon-skew-t distribution, Communications in Statistics Theory and Methods, 36(1), 73-81, 2007
\bibitem{Gomez11} G{\'o}mez, H{\'e}ctor W., Elal-Olivero, D., Salinas H. S., Bolfarine H., Bimodal extension based on the skew-normal distribution with application to pollen data, Environmetrics, 22(1), 50-62, 2011
\bibitem{Goveetal08} Gove J.H., Ducey M.J., Leak W.B., Zhang L. Rotated sigmoid structures in managed uneven-aged northern hardwork stands: a look at the Burr Type
III distribution, Foresty, 81(2), 161-176, 2008
\bibitem{Gui14} Gui W., A generalization of the slashed distribution via alpha skew normal distribution, Statistical Methods \& Applications, 23(1), 1-17, 2014
\bibitem{Hampel86} Hampel F.R., Ronchetti E.M., Rousseeuw P.J., Stahel W.A., Robust Statistics: The Approach based on Influence Functions, John Wiley and Sons, New York, 1986.
\bibitem{hao2009entropy} Hao Z., Singh V.P., Entropy-Based Parameter Estimation for Extended Three-Parameter Burr III Distribution for Low-Flow Frequency Analysis Transactions of the ASABE, 52(4), 1193-1202, 2009
\bibitem{Has10} Hassan Y.M., Hijazi R.H., A Bimodal Exponential Power Distribution, Pak. J. Statist., 26(2):379--396, 2010
\bibitem{Hasel16} Hassan M.Y., El-Bassiouni M.Y., Bimodal skew-symmetric normal distribution, Communications in Statistics - Theory and Methods, 45:5, 1527-1541, 2016
\bibitem{Huber84} Huber P.J., Finite Sample Breakdown of M- and P-Estimators, Annals of Statistics, 12(1), 119-126, 1984
%\bibitem{Hatke49} Hatke M.A. 1949. A certain cumulative probability function, The Annals of Mathematical Statistics, Vol. 20, No. 3, pp.461-463.
\bibitem{Jama11} Jamalizadeh A., Arabpour A. R., Balakrishnan N., Bimodal Class based on the Inverted Symmetrized Gamma Distribution with Applications, Pakistan Journal of Statistics, 52(2), 431-446, 2011
\bibitem{JizbaKorbel16} Jizba P., Korbel J. On q-non-extensive statistics with non-Tsallisian entropy. Physica A: Statistical Mechanics and its Applications, 444, 808-827, 2016
\bibitem{Lindsay96} Lindsay S.R., Wood G.R., Woollons R.C., Modelling the diameter distribution of forest stands using the Burr distribution, 23(6), 609-620, 1996
\bibitem{Lucas97} Lucas A., Robustness of the student t based M-estimator, Commun. Stat. Theory Methods, 26(5), 1165-1182, 1997
\bibitem{mineo2005software} Mineo A.M., Ruggieri M., A software tool for the exponential power distribution: The normalp package, Journal of Statistical Software, 12(4), 1-24, 2005
\bibitem{Mud00} Mudholkar G.S., Hutson A.D., The epsilon-skew-normal distribution for analyzing near-normal data, Journal of statistical planning and inference, 83(2), 291-309, 2000
\bibitem{PurdomHol05} Purdom E., Holmes S.P., Error distribution for gene expression data, Statistical Applications in Genetics and Molecular Biology, 4, Article 16. http://www.bepress.com/sagmb, 2005
\bibitem{Rego12} R{\^e}go L. C., Cintra R. J., Cordeiro G. M., The epsilon-skew-normal distribution for analyzing near-normal data, Journal of Statistical Planning and Inference, 41(20), 3722-3738, 2012
\bibitem{Rocha13} Rocha G. H. M. A., Loschi, R. H., Arellano-Valle, R. B. Inference in flexible families of distributions with normal kernel, Statistics, 47(6), 1184-1206, 2013
\bibitem{Rodriquez77} Rodriguez R. N., A guide to the Burr type XII distributions, Biometrika 64(1), 129-134, 1977
\bibitem{San08} Sanhueza A., Leiva V., Balakrishnan N., The generalized Birnbaum--Saunders distribution and its theory, methodology, and application, Communications in Statistics - Theory and Methods, 37(5), 645-670, 2008
\bibitem{Sham13} Shams H. S., Alamatsaz M. H., Alpha--Skew--Laplace distribution, Statistics \& Probability Letters, 83(3), 774-782, 2013
\bibitem{Shaoetal08} Shao Q. Chen Y.D., Zhang, L. An extension of three-parameter Burr III distribution for low-flow frequency analysis, Computational Statistics Data Analysis 52(3), 1304-1314, 2008
\bibitem{Tadikamalla80} Tadikamalla P.R., A look at the Burr and related distributions, International Statistical Review, 48, 337-344, 1980
%\bibitem{Gomez12} Venegas, Osvaldo and Rodr{\'\i}guez, Francisco and G{\'o}mez, H{\'e}ctor W and Olivares-Pacheco, Juan F and Bolfarine, Heleno Robust modeling using the generalized epsilon-skew-t distribution Journal of Applied Statistics Vol.39(12) 2685-2698
\bibitem{ZhangLi98} Zhang J., Li G., Breakdown properties of location M-estimators, Annals of statistics, 1170-1189, 1998.
\bibitem{zimmer98} Zimmer W. J., Keats J. B., Wang F.K., The Burr XII distribution in reliability analysis, Journal of Quality Technology 30(4), 386-394, 1998
\bibitem{Zoraghietal12} Zoraghi N., Abbasi B., Niaki S.T.A., Abdi M., Estimating the four parameters of the Burr III distribution using a hybrid method of variable neighborhood search and iterated local search algorithms, Applied Mathematics and Computation, 218(19), 9664-9675, 2012
\bibitem{wang96} Wang F.K., Keats J.B., Zimmer W. J., Maximum likelihood estimation of the Burr XII parameters with censored and uncensored data, Microelectronics Reliability, 36(3), 359-362, 1996
\bibitem{Wingo83} Wingo D.R., Maximum Likelihood Methods for fitting the Burr type XII distribution of life test data, Bio. Met. J., 25(1), 77-84, 1983
\bibitem{wingo83} Wingo D. R., Estimating the location of the Cauchy distribution by numerical global optimization, Communications in Statistics-Simulation and Computation, 12(2), 201-212, 1983
%% --- example of Journal article:
%\bibitem{Ab1}
%\textsc{Author, B.}
% (1980).
% Article title in lowercase,
%\textit{Journal Name In Uppercase},
%\textbf{23},
% 4,
% 230--350.
%
%\bibitem{R1}
%\textsc{Rothwell, D.P.}
% (1958).
% Use of varying seasonal weights in price index construction,
%\textit{Journal of the American Statistical Association},
%\textbf{53}, 1, 66--77.
%
%% --- example of Book:
%\bibitem{Ab2}
%\textsc{Author, B.} (1990).
%\textit{Book Name In Uppercase}, Academic Press, New York.
%
%\bibitem{Ro1}
%\textsc{Robert, C.P.} (1999).
%\textit{Monte Carlo Statistical Methods}, Springer Verlag.
%
%% --- example of contribution on Book:
%\bibitem{Ab3}
%\textsc{Author, B.} (2000).
%\textit{Contribution title in lowercase}.
% In "Book Name In Uppercase"
% (J.\ Black and A.\ White, Eds.),
% Academic Press, New York, 123--130.
%
%\bibitem{Ru1}
%\textsc{Runbin, D.G.}
% (1988).
%\textit{Using the SIR algorithm to simulate posterior distributions}.
% In ``Bayesian Statistics"
% (J.M.\ Bernardo, ... and A.F.M.\ Smith, Eds.),
% Oxford University Press, 395--402.
%
%% --- example of 2 authors (book):
%\bibitem{AA1}
%\textsc{Author, B.\ \textup{and} Author, C.} % --- notice the upshape "and"
% (1980).
%\textit{Book Name In Uppercase},
% Academic Press, New York.
%
%% --- example of 3 or more authors (book):
%\bibitem{AAA1}
%\textsc{Author, B.; Author, C.\ \textup{and} Author, D.} % --- ";" and "and"
% (1980).
%\textit{Book Name In Uppercase},
% Academic Press, New York.
\end{thebibliography}
\end{document}