Next Article in Journal
Exact Formulation and Analysis for the Bi-Objective Insular Traveling Salesman Problem
Previous Article in Journal
An Optimal Weighted Combined Model Coupled with Feature Reconstruction and Deep Learning for Multivariate Stock Index Forecasting
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

The Log Exponential-Power Distribution: Properties, Estimations and Quantile Regression Model

1
Department of Measurement and Evaluation, Artvin Çoruh University, City Campus, 08000 Artvin, Turkey
2
Department of Mathematics, Bartin University, 74000 Bartin, Turkey
3
Department of Statistics, Faculty of Intelligent Systems Engineering and Data Science, Persian Gulf University, Bushehr 75169, Iran
4
Department of Mathematics, College of Science and Humanities in Al-Kharj, Prince Sattam bin Abdulaziz University, Al-Kharj 11942, Saudi Arabia
5
Department of Mathematics, Faculty of Science, Mansoura University, Mansoura 35516, Egypt
*
Author to whom correspondence should be addressed.
Mathematics 2021, 9(21), 2634; https://doi.org/10.3390/math9212634
Submission received: 7 August 2021 / Revised: 8 October 2021 / Accepted: 15 October 2021 / Published: 20 October 2021

Abstract

:
Recently, bounded distributions have attracted attention. These distributions are frequently used in modeling rate and proportion data sets. In this study, a new alternative model is proposed for modeling bounded data sets. Parameter estimations of the proposed distribution are obtained via maximum likelihood method. In addition, a new regression model is defined under the proposed distribution and its residual analysis is examined. As a result of the empirical studies on real data sets, it is observed that the proposed regression model gives better results than the unit-Weibull and Kumaraswamy regression models.

1. Introduction

The exponential-power (EP) distribution has been proposed by [1]. If T random variable has EP distribution then its cumulative distribution function (cdf) and probability density function (pdf) are respectively given by
K ( t , α , β ) = 1 e 1 exp α t β
and
k ( t , α , β ) = α β e α t β e α log x β t β 1 e 1 exp α t β ,
where t 0 , α > 0 is the scale parameter and β > 0 is the shape parameter. This distribution is well-known in the literature because of its hazard rate function (hrf) with bathtub shaped. It can be noticed that for β ( 0 , 1 ) , it has hrf with bathtub shaped. Characterizations of the EP distribution was studied by [2].
Several generalizations of the EP distributions have been proposed. For instance, [3] introduced the transmuted-EP (TEP) distribution with applications on the holes operation on jobs made of iron sheet and breaking stress data sets. [4] introduced the generalization of the complementary EP distribution and compared it with other transmuted generalized distributions. [5] introduced the four-parameter generalized EP distribution which has bimodal right-skewed and nearly symmetric pdf shapes.
In the recent years, many lifetime distributions are tranformed to bounded distributions, defined on (0, 1) interval. Two transformation methods are commonly prefered for this goal. These are X = exp T and X = T / T + 1 transformations. When the random variable is defined on T R , the transformed random variable will be T ( 0 , 1 ) . As an example, [6] used X = exp T transformation on the weighted-exponential (WE) distribution and resulting distribution was called as log-WE distribution. Again, [7] used the same transformation for the Bilal distribution and proposed the log-Bilal distribution for extremely left/right skewed data sets defined on the interval ( 0 , 1 ) . Similarly, [8] used X = T / T + 1 transformation on the improved second-degree Lindley (ISDL) distribution and resulting distribution was called as unit-ISDL distribution. These approaches have been used many authors such as 9,10,11,12,13,14,15,16,17,18,19] and so on.
The presented work aims to develop a new sophisticated statistical model for bounded data sets in (0, 1) interval. For this aim, we use X = exp T transformation on the EP distribution. The resulting cdf and pdf of the log-exponential-power (LEP) distribution are given as
F ( x , α , β ) = e 1 exp α log x β , x ( 0 , 1 )
and
f ( x , α , β ) = α β x e α log x β log x β 1 e 1 exp α log x β , x ( 0 , 1 )
respectively, where α > 0 and β > 0 are the model parameters. This new unit model is called as LEP distribution and after here, a random variable X is denoted as X L E P α , β . The associated hrf is given by
h x , α , β = α β x e exp α log x β 1 1 e α log x β log x β 1 , x ( 0 , 1 ) .
If the β parameter is equal to one, then we have following simple cdf and pdf F x , α , 1 = e 1 x α and f x , α , 1 = α x α 1 e 1 x α for x ( 0 , 1 ) respectively. The possible shapes of the pdf and hrf have been sketched by Figure 1. According to this Figure 1, the shapes of the pdf can be seen as various shapes such as U-shaped, increasing, decreasing and unimodal as well as its hrf shapes can be bathtub, increasing and N-shaped.
Other parts of the study are as follows. Statistical properties of the LEP distribution are given in Section 2. Parameter estimation method is presented in Section 3. Section 4 is devoted to the LEP quantile regression model. Section 5 contains two simulation studies for LEP distribution and the LEP quantile regression model. Empirical results of the study are given in Section 6. The study is concluded with Section 7.

2. Some Distributional Properties of the LEP Distribution

The moments, order statistics, entropy and quantile function of the LEP distribution are studied.

2.1. Moments

The n-th non-central moment of the LEP distribution is denoted by E ( X n ) which is defined as
E ( X n ) = n 0 1 x n 1 1 F ( x ) d x = 1 n 0 1 x n 1 e 1 exp ( α ( log ( x ) ) β ) d x
By changing log ( x ) = u transform we obtain
E ( X n ) = 1 + n e 0 e n u e exp ( α u β ) d u = 1 + n e 0 e n u 1 + i = 1 ( 1 ) i i ! exp ( i α u β ) d u = 1 + n e 1 n + i = 1 ( 1 ) i i ! 0 e n u exp ( i α u β ) d u = 1 + e + n e i = 1 j = 0 ( 1 ) i ( α i ) j i ! j ! 0 u β j e n u d u = 1 + e + e i = 1 j = 0 ( 1 ) i ( α i ) j i ! j ! n β j Γ ( β j + 1 )
Based on the first four non-central moments of the LEP distribution, we calculate the skewness and kurtosis values of the LEP distributions. These measures are plotted in Figure 2 against the parameters α and β .

2.2. Order Statistics

The cdf of i-th order statistics of the LEP distribution is given by
F i : n ( x ) = k = 0 n n k F ( x ) k ( 1 F ( x ) ) n k = k = 0 n j = 0 n k ( 1 ) j n k n k j F ( x ) k + j
Then
E ( X i : n r ) = r 0 1 x r 1 1 F i : n ( x ) d x = 1 r k = 0 n j = 0 n k ( 1 ) j n k n k j 0 1 x r 1 e ( k + j ) 1 exp ( α ( log ( x ) ) β ) d x
By changing log ( x ) = u transform we obtain
E ( X i : n r ) = 1 + r k = 0 n j = 0 n k ( 1 ) j n k n k j e k + j 0 e r u e ( k + j ) exp ( α u β ) d u = 1 + r k = 0 n j = 0 n k ( 1 ) j n k n k j e k + j 0 e r u 1 + l = 1 ( 1 ) l ( k + j ) l l ! exp ( l α u β ) d u = 1 + r k = 0 n j = 0 n k ( 1 ) j n k n k j e k + j 1 r + l = 1 s = 0 ( 1 ) l ( k + j ) l ( l α ) s l ! s ! 0 e r u u β s d u = 1 + r k = 0 n j = 0 n k ( 1 ) j n k n k j e k + j 1 r + l = 1 s = 0 ( 1 ) l ( k + j ) l ( l α ) s l ! s ! × Γ ( β s + 1 ) r β s + 1

2.3. Quantile Function and Quantile LEP Distribution

Inverting Equation (3), the quantile function of the L E P distribution is given, we obtain
x τ α , β = e log 1 log τ α 1 / β ,
where τ ( 0 , 1 ) . For the special case β = 1 , we have x τ α , 1 = 1 log τ 1 / α . The median of the L E P distribution has been derived as x 0.5 α , β = exp 0.5266 α 1 / β . Furthermore, if random variable U follows the uniform distribution, the rv X U ( α , β ) L E P ( α , β ) .
On the other hand, since the model has manageable quantile function, its pdf and cdf can be re-parameterized easily. Let define μ = x τ α , β and α = log 1 log τ log μ β . Then the re-parameterized cdf and pdf are obtained as
G y , β , μ , τ = e 1 1 log τ log y / log μ β
and
g y , β , μ , τ = β log 1 log τ 1 log τ log y / log μ β log y β 1 e 1 1 log τ log y / log μ β y log μ β
respectively, where y ( 0 , 1 ) , the μ ( 0 , 1 ) is the quantile parameter, β > 0 is the shape parameter and τ ( 0 , 1 ) is known. Hereafter, the random variable Y will be called quantile LEP (QLEP) random variable and we denote it with Y Q L E P ( β , μ , τ ) . For some selected parameters values, the pdf shapes of the QLEP distribution are displayed in Figure 3. The QLEP distribution has U-shaped, increasing and unimodal shapes.

2.4. Residual Entropy and Cumulative Residual Entropy

Entropy is used to measure uncertainty in different fields such as engineering and natural sciences. The definition of the residual entropy is given by
E ( X ) = 0 F ( x ) log ( F ( x ) ) d x .
The another entropy measure, cumulative residual entropy is defined by
CE ( X ) = 0 F ¯ ( x ) log ( F ¯ ( x ) ) d x .
After some simple algebra, using u = log ( x ) transformation and Taylor expansion for LEP( α , β ) we have
E ( X ) = i , j = 0 1 i α i + 1 j Γ β j + 1 i ! j !
and
CE ( X ) = i , j = 0 e i + 1 ( i + 1 ) j e i + 2 ( i + 2 ) j j ! ( i + 1 ) 1 + k = 1 ( j α ) k Γ ( β k + 1 ) k !

3. Procedure of the Maximum Likelihood for the Parameter Estimation

The maximum likelihood estimators (MLEs) of the LEP distribution has been derived. It has been worked on the case when both α and β are unknown.
Let x 1 , x 2 , , x n random sample of size n from the LEP distribution and let Θ = α , β T be the parameter vector. Then, the log-likelihood function is given by
Θ = n + n log α + n log β + α i = 1 n log x i β + β 1 i = 1 n log log x i i = 1 n e α log x i β .
Then, differentiating (13), the normal equations are obtained by
( Θ ) α = n α + i = 1 n log x i β i = 1 n log x i β e α log x i β = 0
and
( Θ ) β = n β + α i = 1 n log log x i log x i β α i = 1 n log log x i log x i β e α log x i β = 0 .
Above equation systems have no explicit solutions. To obtain the α ^ and β ^ , the numerical approaching should be needed and they have to be solved via numerical methods. The Newton-Raphson and quasi-Newton algorithms can be used for this purpose. However, Equation (13) can be also optimized directly by the special functions in some well-known software such as R (constrOptim and optim and maxLik functions), S-Plus and Matlab. These functions use the numerical optimization methods for solving them. When the log-likelihood is directly optimized, one should carefully choose the initial values and remove the constraints of parameters [20].
The observed information matrix plays an important role in order to obtain the standard errors and asymptotic confidence interval of the MLEs. Based on the regularity conditions, the Θ ^ MLEs have approximately the bivariate normal distribution with mean M = ( α , β ) and covariance matrix I 1 , where I is the observed information matrix with the following elements
I = 2 α 2 ( Θ ) 2 α β ( Θ ) 2 α β ( Θ ) 2 β 2 ( Θ ) Θ = ( α ^ , β ^ ) .
The components of the I can be given by the following derivations
2 α 2 ( Θ ) = n α 2 i = 1 n log x i 2 β e α log x i β ,
2 α β ( Θ ) = i = 1 n log log x i log x i β i = 1 n log log x i log x i β e α log x i β 1 + α log x i β
and
2 β 2 ( Θ ) = n β 2 + α i = 1 n log 2 log x i log x i β α i = 1 n log log x i log log x i log x i β e α log x i β 1 + α log x i β .
The standard errors of the estimated parameters are obtained by the inverse of I, say I 1 . So, the asymptotic confidence intervals of the parameters can be easily constructed by
α ^ ± z ζ / 2 s α ^ and β ^ ± z ζ / 2 s β ^ ,
where s α = I 11 1 , s β = I 22 1 and z ζ / 2 is the upper ζ / 2 th percentile of N 0 , 1 distribution.

4. The New Quantile Regression Model Based on the QLEP Distribution for the Unit Response

The regression models are used to explain the relation between dependent and independent variables. When the dependent variable is on ( 0 , 1 ) interval, the beta regression model by [21] is usually used. The idea of the beta regression model is based on the re-parameterization of the mean and variance functions. Similar idea has been used by several authors such as [6,8,17,22,23]. These studies are constructed on the conditional mean modeling.
The other possible idea is to model the dependent variable with conditional quantile function as in the quantile regression model pioneered by [24]. Ref. [25] introduced the Kumaraswamy quantile regression model for the dependent variables on the ( 0 , 1 ) interval. The presented idea in the study of [25] has been applied to different probability distributions by [9,11,12,14,26,27,28,29].
Following the idea of [25], we propose QLEP quantile regression model. Let Y 1 , Y 2 , , Y n be n random variables from the QLEP distribution denoted by Y Q L E P β , μ i , τ where μ i and β unknown parameters and τ is known. Assume that y 1 , y 2 , , y n are the realizations of the random variables, Y 1 , Y 2 , , Y n . Then, the QLEP regression model is given as follows
g μ i = x i β T ,
where β = β 0 , β 1 , β 2 , , β p T is the unknown regression parameter vector, x i = 1 , x i 1 , x i 2 , x i 3 , , x i p is the known i t h vector of the covariates, and g ( · ) is the link function which connects covariates with conditional quantile of the response variable. The logit link function is preferred by the reason that the response variable is on ( 0 , 1 ) interval. The definition of the logit link function is
g μ i = log μ i 1 μ i .
If the parameter τ = 0.5 , we model the median of the dependent variable with known independent variables which is called as QLEP median regression model.

4.1. The MLEs of the Model Parameters

Let Y 1 , Y 2 , , Y n be n independent random variables from Y i Q L E P β , μ i , τ where
μ i = e x i β T 1 + e x i β T
is the inverse of the logit link function and τ is the known. The Φ = β , β T T be unknown parameter vector with p + 2 . Then, putting (14) in (8), the log-likelihood function of the QLEP regression model is obtained as
Φ = n + n log β + n log log 1 log τ + log 1 log τ i = 1 n log y i log μ i β + β 1 i = 1 n log log y i i = 1 n log y i β i = 1 n log log μ i i = 1 n 1 log τ log y i log μ i β .
The score equations of the Equation (15) are given by
( Φ ) β = n β + log 1 log τ i = 1 n log y i log μ i β log log y i log μ i + i = 1 n log log y i i = 1 n log log μ i log 1 log τ i = 1 n log y i log μ i β log log y i log μ i 1 log τ log y i log μ i β
and
( Φ ) β r = β log 1 log τ i = 1 n log y i log μ i β 1 μ i log μ i μ i β r β i = 1 n 1 μ i log μ i μ i β r + β log 1 log τ i = 1 n log y i log μ i β 1 log τ log y i log μ i β 1 μ i log μ i μ i β r ,
where μ i / β r = μ i 1 μ i x i r , i = 1 , 2 , , n , r = 1 , , p . Above score function consist of the nonlinear function according to model parameters. Therefore, they should be solved by the numerical methods. One may find these solutions via direct maximization of the (15) using software such as R software. The asymptotic distribution of the ( Φ ^ Φ ) is multivariate normal N p + 2 0 , I 1 Φ , where I 1 Φ is the expected information matrix. The p + 2 × p + 2 observed information matrix can be used for the I 1 Φ ^ .

4.2. Model Validity for the Fitting

The model validity is examined via the residual analysis. The randomized quantile residual (rqr) was proposed by [30]. The i t h rqr is calculated by
r ^ i = Φ 1 G ( y i , β ^ , μ ^ i , τ ) ,
where G ( y , β , μ , τ ) is the cdf of the QLEP distribution given by (7), Φ 1 ( x ) is the qf of N ( 0 , 1 ) , and μ ^ i is defined by (14). If the model has good explanatory power, the the rqrs are distributed as the N ( 0 , 1 ) model. The second residual is Cox-Snell residual [31] which is calculated by
e ^ i = log 1 G ( y i , β ^ , μ ^ i , τ ) .
As in rqr, the model is valid, once the e ^ i follows the standard exponential model.

5. Simulation Studies

In this section, we discuss on the simulation studies to see the efficiency of the MLEs of the LEP distribution under different simulation scenarios. Also, we perform another simulation study for the proposed regression model to discuss the asymptotic behaviours of the MLEs of the QLEP quantile regression model. The R software is used to implement the simulation studies.

5.1. Simulation Results for the MLEs of the Proposed Distribution

We focus on the behaviours of the MLEs of the LEP distribution. For this aim, four simulation studies are conducted and the results of these simulation studies are summarized graphically. We have the following settings for the simulation studies. The N = 1000 samples of size n = 20 , 25 , , 1000 from a random variable following the LEP distribution are generated. Four scenarios are considered. The (6) has been used for the random numbers. By setting ϵ = α or β , the related bias and mean square error (MSE) are calculated by
B i a s ϵ n = 1 N i = 1 N ϵ i ϵ ^ i , M S E ϵ n = 1 N i = 1 N ϵ i ϵ ^ i 2 ,
respectively. In addition, for the behaviours of the 95% confidence intervals of the MLEs, we calculate the empirical coverage length (CL) and coverage probability (CP) which are defined by
C L ϵ n = 1 N i = 1 N 3.9198 s ϵ ^ i ,
and
C P h n = 1 N i = 1 N I h ^ i 1.9599 s h ^ i < h < h ^ i + 1.9599 s h ^ i
respectively, where I · is the indicator function and the s ϵ ^ i = s α ^ i , s β ^ i are the standard errors of the MLEs which are calculated by the inverse of the observed information matrix.

5.1.1. Scenario I

The true parameter values for the scenario-I is determined as α = 0.5 and β = 0.5 . The estimated means, MSEs and biases for each sample size are displayed in Figure 4. The estimated CLs and CPs are also displayed in Figure 5. From these results, we conclude that the estimated biases and MSEs are near the zero. The estimated means approach the true values of each parameters. As expected, the CPs are near the desired value, 0.95 for all sample sizes. The CLs decrease when the sample sizes increase. In the light of these results, it is verified that the consistency property of the MLE is valid for the LEP distribution.

5.1.2. Scenario II

The true parameters are determined as α = 2 and β = 2 for the scenario-II and the results are summarized in Figure 6 and Figure 7. Since the results of the scenario-II is the same with scenario-I, the interpretation of the simulation results are omitted. These results also verify the suitability of the MLE method for the LEP distribution.

5.1.3. Scenario III

The true parameter values for the scenario-III is determined as α = 0.5 and β = 2 . The simulation results are displayed in Figure 8 and Figure 9. The results are similar with the results of the scenario-I. Therefore, the interpretation of the results are omitted. As in previous simulation studies, these results verify the suitability of the MLE method for the proposed distribution.

5.1.4. Scenario IV

For the last scenario, the true values of the parameters are determined as α = 2 and β = 0.5 . Figure 10 and Figure 11 give the results of the simulation study graphically. General result of these simulation studies is that the MLE method works well to estimate the unknown parameter of the LEP distribution.

5.2. Comparison of SD and SE

Here, we compare the average of the standard errors (SEs) and standard deviations (SDs) of the estimated parameters to evaluate the unbiasedness of the SEs. For this aim, the SDs and SEs are calculated for four different scenarios and the results are graphically summarized in Figure 12, Figure 13, Figure 14 and Figure 15. If the SEs are unbiased, we expect to see that SDs and SEs should be near to each other. As seen from Figure 12, Figure 13, Figure 14 and Figure 15, the values of the SDs and SEs are near to each other. The simulation results verify the unbiasedness of the SEs.

5.3. Simulation Studies for the Proposed Regression Model

We give the simulation study in order to evaluate MLEs of the parameters of the QLEP regression model based on the bias and MSE calculations. For the varying sample size n, known τ , true β , and generated covariates values, the values of the unit response variable have been obtained with
y i = exp log 1 log u i α i 1 / β ,
where u i U ( 0 , 1 ) and α i = log 1 log τ log μ i β for i = 1 , 2 , , n .
In this simulation study, it has been considered as replication number N = 1000 , sample sizes n = 25 , 50 , 100 , 250 , 500 , τ = 0.25 , 0.50 , 0.75 , β = 5 with the following regression structure
logit ( μ i ) = β 0 + β 1 z i 1 , i = 1 , 2 , , n ,
where β 0 = 0.5 , β 1 = 1 and z i 1 N ( 0 , 1 ) .
Table 1 shows the simulation results for the QLEP regression model. As seen from these results, the empirical CLs decrease while the sample size increases as well as the empirical CPs are around the 0.95 value. All biases are close zero value as wel as all MSEs tend to zero value at the same time.

6. Applications

The presented section aims to show the importance of the modeling ability of both LEP distribution and QLEP regression model, based on the real data sets. The maximized log-likelihood value ^ , Akaike Information Criteria (AIC), Bayesian information criterion (BIC), Cramer-von-Mises, ( W * ), Anderson-Darling ( A * ) and Kolmogorov-Smirnov (KS) goodness of-fit statistics have been calculated based on all distribution models to choose the optimum model. It is chosen as the optimum model the one which has the smaller the values of the AIC, BIC, W * and A * statistics and the larger the values of ^ . The calculations have been obtained by the maxLik and goftest functions of the R software. The details are in the following subsections.

6.1. Real Data Application for the Univariate Data Modeling

The data includes in the flood levels for the Susquehanna River at Harrisburg, Pennsylvania, and represents the maximum flood level (in millions of cubic feet per second). These data are: 0.654, 0.613, 0.315, 0.449, 0.297, 0.402, 0.379, 0.423, 0.379, 0.3235, 0.269, 0.740, 0.418, 0.412, 0.494, 0.416, 0.338, 0.392, 0.484, 0.265. Recenly, this data set has been analyzed by [32,33].
Well known unit distributions in the literature have been compared with our distribution under the MLE method. Their densities have been given as follows.
  • Beta distribution:
    f B e t a ( x , α , β ) = 1 B ( α , β ) x α 1 1 x β 1 , x ( 0 , 1 ) ,
    and f B e t a ( x , α , β ) = 0 for x ( 0 , 1 ) , where α > 0 , β > 0 , and B ( α , β ) is the standard beta function.
  • Kumaraswamy (Kw) distribution [34]:
    f K w ( x , α , β ) = α β x α 1 1 x α β 1 , x ( 0 , 1 ) ,
    and f K w ( x , α , β ) = 0 for x ( 0 , 1 ) , where α > 0 and β > 0 .
  • Johnson S B distribution [35]:
    f S B ( x , α , β ) = β x 1 x ϕ β log x 1 x + α , x ( 0 , 1 ) ,
    and f S B ( x , α , β ) = 0 for x ( 0 , 1 ) , where α R , β > 0 , and the ϕ ( x ) is the pdf of the standard normal distribution.
  • Unit Birnbaum Saunders ( U B S ) distribution [36]:
    f U B S x , α , β = 1 2 x α β 2 π β log x + β log x 3 e 1 2 α 2 log x β + β log x + 2 ,
    and f U B S ( x , α , β ) = 0 for x ( 0 , 1 ) , where α > 0 and β > 0 .
Table 2 shows the results of data analysis. The standard errors are in · and p-values of the KS tests are in · From this Table, we can say that the LEP distribution is the best model based on above comparison criteria.
Figure 16 displays the estimated pdfs and cdfs of all distributions according to results of the fitted models. The probability-probability (PP) Quantile-Quantile (QQ) plots of the fitted LEP distribution have been given by Figure 17. All these plots indicates the acceptable fitting of the LEP distribution for the data set.
The profile log-likelihood (PLL) functions are plotted in Figure 18 for the parameters of the QLEP distribution. According to Figure 18, the estimated parameters are maximizers of the function in (13).

6.2. Quantile Regression Application

We use the Better Life Index (BLI) data set measured in the year of 2017, available at https://stats.oecd.org/index.aspx?DataSetCode=BLI (accessed on 7 June 2021), to demonstrate the practicality of the QLEP regression model. We compare the QLEP model with two quantile regression models: Kumaraswamy and unit-Weibull. Homicide rate, y i , is considered as dependent variable. The covariates are: employment rate x i 1 , personal earnings x i 2 and labour market insecurity x i 3 . The goal of the presented application is to explain the variability of the homicide rate with these covariates.
The estimated parameters as well as model selection criteria of the quantile regression models for τ = 0.5 are given in Table 3. The calculated AIC and BIC values show that the QLEP regression model is better than the Kumaraswamy and unit-Weibull regression models for the considered data set since the proposed model has the lowest value of these statistics. According to estimated regression parameters, β 2 is found statistically significant at % 5 level for three regression models. It means that when the personal earnings increase, the homicide rate decreases. It is an expected outcome. Because, the countries having high earnings provide better life conditions and it decreases the homicide rate. It is widely documented that the income inequality leads to increased homicide rate [37].
Accuracy of the fitted regression models are evaluated by means of the rqrs. As mentioned before, when the fitted model is preferable, the rqrs should be distributed as N 0 , 1 . Figure 19 shows the QQ plots of the rqrs for all fitted regression models. From these figures, one can conclude that the QLEP model is more appropriate model than others because the plotted points for the LEP regression models is more closer the diagonal line than those of others.
Also, Table 4 lists the KS test results to verify that whether the randomized quantile residuals are distributed as standard normal distribution. Obviously, all p-values are higher than 0.05 . So, randomized quantile residuals are normally distributed for all regression models. However, the p-value of the QLEP regression model is higher than those of other models. It is also evidence for the superiority of the QLEP regression model over two other models.

7. Conclusions

In this study, a quantile regression model is defined under the proposed distribution. The parameter estimates of the proposed regression model are obtained by the maximum likelihood estimation method, and the efficiency of the estimation method is examined via simulation study. Homicide rate of the OECD countries are analyzed by the proposed approach as well as unit-Weibull and Kumaraswamy regression models. The residual analysis of the fitted regression models are performed with the randomized quantile residuals. Based on the residual analysis and model selection criteria, the proposed approach is selected as a best model among others. Additionally, the distance-based model selection criteria is used in the study. However, the likelihood-based model selection methods are more effective than the distance-based methods [38]. So, the comparison of the LEP model with existing bounded distributions can be performed under the likelihood-based model selection approaches. We plan this issue as a future work of the presented study. Moreover, we plan to develop an extension of the proposed model when the response variable is time-dependent such as monthly incidence ratio of the corona-virus disease or weekly car accident involving death ratio.

Author Contributions

Conceptualization, M.Ç.K., E.A., M.A. and M.E.-M.; methodology, M.Ç.K., E.A. and M.A.; software, M.Ç.K. and E.A.; validation, M.Ç.K., E.A., M.A. and M.E.-M.; formal analysis, M.Ç.K., E.A. and M.A.; investigation, M.Ç.K., E.A., M.A. and M.E.-M.; resources, M.Ç.K., E.A., M.A. and M.E.-M.; data curation, M.Ç.K., E.A. and M.E.-M.; writing—original draft preparation, M.Ç.K., E.A., M.A. and M.E.-M.; writing—review and editing, M.Ç.K., E.A., M.A. and M.E.-M.; visualization, M.Ç.K. and E.A. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

The data sources are presented in Section 6.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Smith, R.M.; Bain, L.J. An exponential power life-testing distribution. Commun. Stat. Theory Methods 1975, 4, 469–481. [Google Scholar]
  2. Shakil, M.; Ahsanullah, M.; Kibria, B.G. On the Characterizations of Chen’s Two-Parameter Exponential Power Life-Testing Distribution. J. Stat. Theory Appl. 2018, 17, 393–407. [Google Scholar]
  3. Saraçoğlu, B.; Taniş, C. A new lifetime distribution: Transmuted exponential power distribution. Commun. Fac. Sci. Univ. Ank. Ser. Math. Stat. 2021, 70, 1–14. [Google Scholar]
  4. Taniş, C.; Saraçoğlu, B.; Coşkun, K.; Pekgör, A. Transmuted complementary exponential power distribution: Properties and applications. Cumhur. Sci. J. 2020, 41, 419–432. [Google Scholar] [CrossRef]
  5. Abd El-Monsef, M.; El-Awady, M. Generalized Exponential Power Distribution with Application to Complete and Censored Data. Asian J. Probab. Stat. 2021, 34–55. [Google Scholar] [CrossRef]
  6. Altun, E. The log-weighted exponential regression model: Alternative to the beta regression model. Commun. Stat. Theory Methods 2021, 50, 2306–2321. [Google Scholar] [CrossRef]
  7. Altun, E.; El-Morshedy, M.; Eliwa, M.S. A new regression model for bounded response variable: An alternative to the beta and unit-Lindley regression models. PLoS ONE 2021, 16, e0245627. [Google Scholar] [CrossRef]
  8. Altun, E.; Cordeiro, G.M. The unit-improved second-degree Lindley distribution: Inference and regression modeling. Comput. Stat. 2020, 35, 259–279. [Google Scholar] [CrossRef]
  9. Korkmaz, M.Ç.; Emrah, A.; Chesneau, C.; Yousof, H.M. On the unit-Chen distribution with associated quantile regression and applications. Math. Solovaca 2021. accepted paper. [Google Scholar]
  10. Altun, E.; Hamedani, G. The log-xgamma distribution with inference and application. J. Soc. Française Stat. 2018, 159, 40–55. [Google Scholar]
  11. Mazucheli, J.; Leiva, V.; Alves, B.; Menezes, A.F.B. A new quantile regression for modeling bounded data under a unit Birnbaum–Saunders distribution with applications in medicine and politics. Symmetry 2021, 13, 682. [Google Scholar] [CrossRef]
  12. Korkmaz, M.Ç.; Chesneau, C.; Korkmaz, Z.S. On the Arcsecant Hyperbolic Normal Distribution. Properties, Quantile Regression Modeling and Applications. Symmetry 2021, 13, 117. [Google Scholar] [CrossRef]
  13. Bakouch, H.S.; Nik, A.S.; Asgharzadeh, A.; Salinas, H.S. A flexible probability model for proportion data: Unit-Half-Normal distribution. Commun. Stat. Case Stud. Data Anal. Appl. 2021, 7, 271–288. [Google Scholar]
  14. Korkmaz, M.Ç.; Chesneau, C. On the unit Burr-XII distribution with the quantile regression modeling and applications. Comput. Appl. Math. 2021, 40, 29. [Google Scholar] [CrossRef]
  15. Korkmaz, M.Ç. The unit generalized half normal distribution: A new bounded distribution with inference and application. Appl. Math. Phys. 2020, 2, 133–140. [Google Scholar]
  16. Mazucheli, J.; Bapat, S.R.; Menezes, A.F.B. A new one-parameter unit-Lindley distribution. Chil. J. Stat. 2019, 11, 53–67. [Google Scholar]
  17. Gómez-Déniz, E.; Sordo, M.A.; Calderín-Ojeda, E. The Log-Lindley distribution as an alternative to the beta regression model with applications in insurance. Insur. Math. Econ. 2014, 54, 49–57. [Google Scholar] [CrossRef]
  18. Mazucheli, J.; Menezes, A.F.B.; Ghitany, M.E. The unit-Weibull Distribution and associated inference. J. Appl. Probab. Stat. 2018, 13, 1–22. [Google Scholar]
  19. Mazucheli, J.; Menezes, A.F.; Dey, S. Unit-Gompertz distribution with applications. Statistica 2019, 79, 25–43. [Google Scholar]
  20. MacDonald, I.L. Does Newton–Raphson really fail? Stat. Methods Med. Res. 2014, 23, 308–311. [Google Scholar] [CrossRef]
  21. Ferrari, S.; Cribari Neto, F. Beta regression for modelling rates and proportions. J. Appl. Stat. 2004, 31, 799–815. [Google Scholar] [CrossRef]
  22. Bayes, C.L.; Bazán, J.L.; García, C. A new robust regression model for proportions. Bayesian Anal. 2012, 7, 841–866. [Google Scholar] [CrossRef]
  23. Mazucheli, J.; Menezes, A.F.B.; Chakraborty, S. On the one parameter unit-Lindley distribution and its associated regression model for proportion data. J. Appl. Stat. 2019, 46, 700–714. [Google Scholar] [CrossRef] [Green Version]
  24. Koenker, R.; Bassett, G.J. Regression quantiles. Econ. J. Econ. Soc. 1978, 46, 33–50. [Google Scholar] [CrossRef]
  25. Mitnik, P.A.; Baek, S. The Kumaraswamy distribution: Median-dispersion re-parameterizations for regression modeling and simulation-based estimation. Stat. Pap. 2013, 54, 177–192. [Google Scholar] [CrossRef]
  26. Bayes, C.L.; Bazán, J.L.; Castro, M. A quantile parametric mixed regression model for bounded response variables. Stat. Its Interface 2017, 10, 483–493. [Google Scholar] [CrossRef]
  27. Mazucheli, J.; Menezes, A.F.B.; Fernandes, L.B.; de Oliveira, R.P.; Ghitany, M.E. The unit-Weibull distribution as an alternative to the Kumaraswamy distribution for the modeling of quantiles conditional on covariates. J. Appl. Stat. 2020, 47, 954–974. [Google Scholar] [CrossRef]
  28. Jodrá, P.; Jiménez-Gamero, M.D. A quantile regression model for bounded responses based on the Exponential-Geometric distribution. Revstat-Stat. J. 2020, 4, 415–436. [Google Scholar]
  29. Korkmaz, M.; Chesneau, C.; Korkmaz, Z. Transmuted unit Rayleigh quantile regression model: Alternative to beta and Kumaraswamy quantile regression models. Univ. Politeh. Buchar. Sci. Bull. Ser. Appl. Math. Phys. 2021, 83, 149–158. [Google Scholar]
  30. Dunn, P.K.; Smyth, G.K. Randomized quantile residuals. J. Comput. Graph. Stat. 1996, 5, 236–244. [Google Scholar]
  31. Cox, D.R.; Snell, E.J. A general definition of residuals. J. R. Stat. Soc. Ser. B (Methodol.) 1968, 30, 248–265. [Google Scholar] [CrossRef]
  32. Balakrishnan, N.; Cohen, A.C. Order Statistics & Inference: Estimation Methods; Elsevier: Amsterdam, The Netherlands, 2014. [Google Scholar]
  33. Alizadeh, M.; Altun, E.; Cordeiro, G.M.; Rasekhi, M. The odd power cauchy family of distributions: Properties, regression models and applications. J. Stat. Comput. Simul. 2018, 88, 785–807. [Google Scholar] [CrossRef]
  34. Kumaraswamy, P. A generalized probability density function for double-bounded random processes. J. Hydrol. 1980, 46, 79–88. [Google Scholar] [CrossRef]
  35. Johnson, N.L. Systems of frequency curves generated by methods of translation. Biometrika 1949, 36, 149–176. [Google Scholar] [CrossRef]
  36. Mazucheli, J.; Menezes, A.F.B.; Dey, S. The unit-Birnbaum-Saunders distribution with applications. Chil. J. Stat. 2018, 9, 47–57. [Google Scholar]
  37. Ouimet, M. A world of homicides: The effect of economic development, income inequality, and excess infant mortality on the homicide rate for 165 countries in 2010. Homicide Stud. 2012, 16, 238–258. [Google Scholar] [CrossRef]
  38. Mitra, D.; Kundu, D.; Balakrishnan, N. Likelihood analysis and stochastic EM algorithm for left truncated right censored data and associated model selection from the Lehmann family of life distributions. Jpn. J. Stat. Data Sci. 2021, 1–30. [Google Scholar]
Figure 1. The possible shapes of the pdf (left) and hrf (right).
Figure 1. The possible shapes of the pdf (left) and hrf (right).
Mathematics 09 02634 g001
Figure 2. The skewness (left) and kurtosis (right) plots of LEP distribution.
Figure 2. The skewness (left) and kurtosis (right) plots of LEP distribution.
Mathematics 09 02634 g002
Figure 3. The pdf shapes of the QLEP distribution.
Figure 3. The pdf shapes of the QLEP distribution.
Mathematics 09 02634 g003
Figure 4. The results of the α (top) and β (bottom) parameters for the simulation-I.
Figure 4. The results of the α (top) and β (bottom) parameters for the simulation-I.
Mathematics 09 02634 g004
Figure 5. The empirical CLs and CPs of the MLEs for the scenario-I.
Figure 5. The empirical CLs and CPs of the MLEs for the scenario-I.
Mathematics 09 02634 g005
Figure 6. The results of the α (top) and β (bottom) parameters for the simulation-II.
Figure 6. The results of the α (top) and β (bottom) parameters for the simulation-II.
Mathematics 09 02634 g006
Figure 7. The empirical CLs and CPs of the MLEs for the scenario-II.
Figure 7. The empirical CLs and CPs of the MLEs for the scenario-II.
Mathematics 09 02634 g007
Figure 8. The results of the α (top) and β (bottom) parameters for the simulation-III.
Figure 8. The results of the α (top) and β (bottom) parameters for the simulation-III.
Mathematics 09 02634 g008
Figure 9. The empirical CLs and CPs of the MLEs for the scenario-III.
Figure 9. The empirical CLs and CPs of the MLEs for the scenario-III.
Mathematics 09 02634 g009
Figure 10. The results of the α (top) and β (bottom) parameters for the simulation-IV.
Figure 10. The results of the α (top) and β (bottom) parameters for the simulation-IV.
Mathematics 09 02634 g010
Figure 11. The empirical CLs and CPs of the MLEs for the scenario-IV.
Figure 11. The empirical CLs and CPs of the MLEs for the scenario-IV.
Mathematics 09 02634 g011
Figure 12. Comparison of SD and SE for the scenario-I.
Figure 12. Comparison of SD and SE for the scenario-I.
Mathematics 09 02634 g012
Figure 13. Comparison of SD and SE for the scenario-II.
Figure 13. Comparison of SD and SE for the scenario-II.
Mathematics 09 02634 g013
Figure 14. Comparison of SD and SE for the scenario-III.
Figure 14. Comparison of SD and SE for the scenario-III.
Mathematics 09 02634 g014
Figure 15. Comparison of SD and SE for the scenario-IV.
Figure 15. Comparison of SD and SE for the scenario-IV.
Mathematics 09 02634 g015
Figure 16. Estimated pdfs (left) and cdfs (right) for the application models.
Figure 16. Estimated pdfs (left) and cdfs (right) for the application models.
Mathematics 09 02634 g016
Figure 17. The PP (left) and QQ plots (right) of the LEP distribution for the fitted model.
Figure 17. The PP (left) and QQ plots (right) of the LEP distribution for the fitted model.
Mathematics 09 02634 g017
Figure 18. The plots of the PLL functions.
Figure 18. The plots of the PLL functions.
Mathematics 09 02634 g018
Figure 19. QQ plots of the rqrs: (left) QLEP, (middle) Kumaraswamy an (right) unit-Weibull.
Figure 19. QQ plots of the rqrs: (left) QLEP, (middle) Kumaraswamy an (right) unit-Weibull.
Mathematics 09 02634 g019
Table 1. The empirical biases, MSE, 95 % CP, and 95 % CL for the QLEP regression model with τ = 0.25 , 0.50 and 0.75 .
Table 1. The empirical biases, MSE, 95 % CP, and 95 % CL for the QLEP regression model with τ = 0.25 , 0.50 and 0.75 .
τ = 0.25
nBias- β 0 Bias- β 1 Bias- β MSE- β 0 MSE- β 1 MSE- β CL- β 0 CL- β 1 CL- β CP- β 0 CP- β 1 CP- β
25 0.0100 0.0020 0.4837 0.0013 0.0021 1.3551 0.1255 0.1636 3.7139 0.8900 0.9250 0.9440
50 0.0030 0.0023 0.2659 0.0007 0.0006 0.5236 0.0957 0.0894 2.4852 0.9300 0.9290 0.9430
100 0.0028 0.0017 0.0914 0.0003 0.0003 0.1972 0.0671 0.0614 1.6859 0.9420 0.9440 0.9560
250 0.0003 0.0007 0.0360 0.0001 0.0001 0.0728 0.0430 0.0401 1.0470 0.9470 0.9460 0.9530
500 0.0001 0.0005 0.0227 0.0001 0.0001 0.0385 0.0303 0.0293 0.7375 0.9480 0.9390 0.9460
τ = 0.5
nBias- β 0 Bias- β 1 Bias- β MSE- β 0 MSE- β 1 MSE- β CL- β 0 CL- β 1 CL- β CP- β 0 CP- β 1 CP- β
25 0.0007 0.0066 0.4996 0.0020 0.0020 1.2113 0.1713 0.1595 3.7157 0.9350 0.9100 0.9610
50 0.0003 0.0015 0.2131 0.0011 0.0008 0.4604 0.1264 0.1134 2.4639 0.9410 0.9450 0.9550
100- 0.0011 0.0013 0.1285 0.0006 0.0003 0.2129 0.0920 0.0696 1.6977 0.9440 0.9280 0.9530
250- 0.0008 0.0006 0.0531 0.0002 0.0001 0.0742 0.0582 0.0439 1.0503 0.9490 0.9610 0.9570
500 0.0003 0.0002 0.0273 0.0001 0.0001 0.0363 0.0413 0.0301 0.7368 0.9400 0.9500 0.9470
τ = 0.75
nBias- β 0 Bias- β 1 Bias- β MSE- β 0 MSE- β 1 MSE- β CL- β 0 CL- β 1 CL- β CP- β 0 CP- β 1 CP- β
25- 0.0117 0.0038 0.5318 0.0057 0.0016 1.4534 0.2782 0.1340 3.7260 0.8990 0.9050 0.9430
50 0.0069 0.0050 0.2627 0.0026 0.0008 0.5056 0.1984 0.1017 2.4788 0.9340 0.9010 0.9440
100− .0025 0.0018 0.0973 0.0013 0.0003 0.1987 0.1440 0.0700 1.6731 0.9430 0.9450 0.9480
250− .0018 0.0009 0.0533 0.0005 0.0001 0.0714 0.0913 0.0406 1.0400 0.9500 0.9470 0.9510
500 0.0015 0.0003 0.0310 0.0003 0.0001 0.0378 0.0644 0.0317 0.7336 0.9400 0.9480 0.9460
Table 2. The results of the fitted models for the maximum flood level data.
Table 2. The results of the fitted models for the maximum flood level data.
Model α ^ β ^ ^ AIC BIC A * W * KS p - Value
LEP0.65932.919416.5800−29.1593−28.45340.29230.04980.1366 0.8494
(0.1128)(0.5537)
Johnson S B 0.61431.926114.2629−24.5257−22.53420.69320.11540.1935 0.4421
(0.2438)(0.3035)
Beta6.75689.111414.0622−24.1244−22.13300.73270.12360.1988 0.4082
(2.7299)(3.4697)
U B S 0.37830.837412.6784−21.3568−19.36531.05160.18980.2286 0.2470
(0.0598)0.0696
Kw3.363211.789212.8662−21.7324−19.74090.93220.16360.2109 0.3359
(0.6143)(5.4906)
Table 3. Estimated parameters and model selection criteria of the fitted quantile regression models.
Table 3. Estimated parameters and model selection criteria of the fitted quantile regression models.
ParametersQLEPKumaraswamyUnit-Weibull
EstimatesSEsp-ValuesEstimatesSEsp-ValuesEstimatesSEsp-Values
β 0 −1.7942.0970.1960.1492.4840.476-2.9382.3510.106
β 1 −0.0250.0310.208−0.0340.0350.167-0.0040.0340.456
β 2 −0.0260.0140.031−0.0480.011<0.001-0.0340.0150.010
β 3 −0.0040.0400.463−0.0480.0510.1740.0150.0460.369
β 4.4260.661-1.0040.121-5.6250.778-
AIC−223.970−209.571−219.351
BIC−215.782−201.383−211.163
Table 4. KS test results for the rqrs of the models.
Table 4. KS test results for the rqrs of the models.
KSQLEPKumaraswamyUnit-Weibull
Test statistic0.0970.1480.102
p-value0.8240.3350.782
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Share and Cite

MDPI and ACS Style

Korkmaz, M.Ç.; Altun, E.; Alizadeh, M.; El-Morshedy, M. The Log Exponential-Power Distribution: Properties, Estimations and Quantile Regression Model. Mathematics 2021, 9, 2634. https://doi.org/10.3390/math9212634

AMA Style

Korkmaz MÇ, Altun E, Alizadeh M, El-Morshedy M. The Log Exponential-Power Distribution: Properties, Estimations and Quantile Regression Model. Mathematics. 2021; 9(21):2634. https://doi.org/10.3390/math9212634

Chicago/Turabian Style

Korkmaz, Mustafa Ç., Emrah Altun, Morad Alizadeh, and M. El-Morshedy. 2021. "The Log Exponential-Power Distribution: Properties, Estimations and Quantile Regression Model" Mathematics 9, no. 21: 2634. https://doi.org/10.3390/math9212634

APA Style

Korkmaz, M. Ç., Altun, E., Alizadeh, M., & El-Morshedy, M. (2021). The Log Exponential-Power Distribution: Properties, Estimations and Quantile Regression Model. Mathematics, 9(21), 2634. https://doi.org/10.3390/math9212634

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop