Exponentiated Discrete Hypo Exponential Distribution and its Generalizations Текст научной статьи по специальности «Науки о Земле и смежные экологические науки»

Exponentiated Discrete Hypo Exponential Distribution

and its Generalizations

Krishnakumari.K1 and Dais George2

Research Scholar, St.Thomas College, Palai (SAS SNDP YOGAM College, Konni), Kerala, India. 2Catholicate College, Pathanamthitta, Kerala, India.

krishnavidyadharan@gmail.com, daissaji@rediffmail.com

Abstract

Generalizations of standard probability distributions is a thought-provoking concept in statistical literature and was inspired by many researchers in recent days. This is because the addition of parameters may increase the flexibility of the new models. Now a days various generalization techniques are available in literature. In this work, we proposed a generalization of discrete hypo exponential distribution and studied its various properties. A real data analysis is carried out and check the flexibility of the new model by comparing it with other standard distributions. Two generalizations of the proposed distribution are introduced.

Keywords: Discrete hypo exponential distribution, Estimation, Generalization, Moments, Stress-strength analysis.

1. Introduction

Over the last few decades, there has been growing interest in adding supplementary parameters to the baseline distributions to broaden generalized families of distributions. The addition of parameters may increase the flexibility of the new models. So generalization of the standard distributions are attracted by many researchers and are prominent in recent days. In literature there exists various generalization techniques and for a detailed review, see Tahir and Nadarajah[32]. These techniques resulted in the generalizations of various standard distributions. For details see, Gupta and Kundu [14], Eugene et.al [11], Zografos and Balakrishnan [33], Gomez-Deniz [13], Mahmoudi and Zakerzadeh [19], Cordeiro and Castro [4], Nadarajah [24], Nadarajah et.al [26], Cordeiro et al. [5], Ristic and Balakrishnan [30], Lemonte et.al [17], Liyanage and Pararai [18], Merovci and Elbatal [22], Merovci and Sharma [23], Nadarajah and Bakar [25], Ahmad and Ghazal [1], Sulami [2] etc. Recently exponentiated family of distributions due to Lehman [16] has got special attention and various standard distributions were generalized. The most prominent distribution introduced in the 20th century is the exponentiated exponential distribution and inspired by this many existing distributions were generalized and for details see, Pal et al. [28], Nekoukhou and Bidram [27], Morshedy et al. [8], El-Bassiouny et al. [7], Morshedy et al.[10], Morshedy et al. [9], Mashhadzadeh and Mirmostafaee [21] and Baharith and Alamoudi [3]. The layout of this article is in this way. In Section 2, we introduced exponentiated discrete hypo exponential distribution and studied its various properties. In Section 3 the parameters of the distribution is done through non linear maximization method. To evaluate the performance of the nlm estimator a simulation study is done in Section 4. A real data analysis is done in Section 5. In Section 6 some generalizations of the proposed distribution are introduced. Some concluding remarks are recorded in Section 7.

2. Exponentiated Discrete Hypo Exponential Distribution

Consider the discrete hypo exponential (DHE) distribution having model parameters 01, 02 > 0, _ 02 with the distribution function

F(x; 01,02)

- 01

(1

_ e-01x) _

- 01

(1

_ e_02x

)

(1)

By inserting (1) into the resilience parameter family of distributions, the distribution function of the resulting distribution is given by

where

G(x; 01,02, a) _ [F(x; 01,02 )]a _ V(x; 01,02, a) _ (02 _ 01 )a

V(x; 01,02, a) _ [02(1 _ e_01 x) _ 01(1 _ e_02x)]

(2)

(3)

We call such a random variable X, having distribution function (2), is an exponentiated DHE distribution with parameters 01, 02 > 0, 01 _ 02, a > 0 and denote it as EDHE (01,02, a). The probability mass function(pmf) of EDHE distribution is given by

P(X _ x) _ v(x; 01,02, a)

V(x + 1; 01,02, a) _ V(x; 01,02, a) _ (02 _ 01)a The plots of pmf of EDHE distribution is given in Figure 1.

;x _ 0,1,2,....

(4)

Figure 1: Plots of pmf of EDHE distribution

From Figure 1 it is understood that the EDHE distribution is unimodel. Since every log-concave density is unimodel, it is also inferred that EDHE distribution is log-concave.

a

2.1. Reliability characteristics

Survival function, S(x) = 1 — Vf1, f2'^ ; x = 0,1,2..., and

hazard rate, r(x)

(f 2 — f )Й V (x + 1; f 1, f 2, a) — V (x; f 1, f 2, a)

(f 2 — f1)a — V (x; f1, f 2, a) The plots of hazard rate of EDHE distribution is given in Figure 2.

Figure 2: Plots of hazard rate of EDHE distribution.

From Figure 2, it is evident that for various model parameters, the hazard rate functions can be decreasing, increasing and increasing-decreasing, which makes the EDHE distribution more flexible and can model different types of data sets such as count data, failure time data etc.

2.2. Moments

Let X - EDHE(f 1, f2, a), then for n > 1,

V (x + 1; fi, f2, a) — V (x; fi, f2, a)

TO

E(Xn) = £ xn

x=0

In particular

and

E(X)= £ x

x=0

TO

E(X2) = £ x2

x=0

(f 2 — f1)a V (x + 1; f 1, f 2, a) — V(x; f 1, f2, a)

V (x + 1; f1, f2, a) — V (x; f1, f2, a)

The expression for V(X) can be obtained using he relation

V(X) = E(X2) - [E(X)]2

2.3. Infinite Divisibility

According to Steutel and Van Harn [31], a necessary condition for infinite divisibility of a discrete distribution Py is that Po > 0. For EDHE distribution this condition is satisfied for all values of the parameters. Hence it is infinitely divisible.

2.4. Theorem

If X follows an exponentiated hypo exponential distribution with parameters , 02 and a then the random variable W=[X] follows a exponentiated discrete hypo exponential distribution with parameters 01, 02 and a .

TO

Proof:

Consider w=0,1, 2,... then using Lemmal of Krishna and Pundir (2009), we have

P(W > w) = P([X] > w) = P(X > w)

1

02(1 - e-01x) - 01 (1 - e-02x)

02 - 01

which is the survival function of EDHE distribution and hence the theorem.

2.5. Order Statistics

Order statistics are sample values placed in ascending order. It is a very useful concept in statistical sciences. It has a far reaching applications especially in modeling auctions, car races, insurance policies and estimating parameters of distributions etc.

Let X(1:n) < X(2:n) < X(3:n) < ... < X(n.n) represents the order statistics obtained from the i.i.d. EDHE(01,02, a) distribution of size n. Then probability mass function of first order statistics is given by

fX(1:n)(x)

1

V(x - 1; 01,02, a)

(02 - 01 )a

1

V(x; 01,02, a) (02 - 01 )a

and the distribution function is

FX(1:n)(x) = 1 -

1

V (x; 01,02, a) (02 - 01)a

The probability mass function of nth order statistics is given by

fX(nn)(x) =

and the distribution function is

V(x; 01,02, a) (02 - 01)a

V(x - 1; 01,02, a)

(02 - 01)a

FX ( x)

(n:n)

V(x; 01,02, a) (02 - 01 )a

where V(x; 01,02, a) is given by (3).

2.6. Entropy

The Shannon's entropy of random variable X having probability mass function P(x) is given by

H(X) = E(-logP(x)).

For EDHE distribution, H(X) is obtained as

H(X) = - £

x=0

V(x + 1; 01,02, a) - V(x; 01,02, a) (02 - 01)a

log

V(x + 1; 01,02, a) - V(x; 01,02, a) (02 - 01)a

Renyi's entropy of order fi > 0 (fi = 1) is

Hp (p)

1

1- fil°g £ 1 fi x=0

V(x + 1; 01,02, a) - V(x; 01,02, a) (02 - 01 )a

a

n

n

n

n

n

n

fi

CO

2.7. Stress-strength Analysis

Stress-strength analysis is a mechanism and a topic used in reliability engineering. When the probability of stress exceeding the strength of an item, that item fails. Hence the expected reliability (R) is calculated as

TO

R = P (stress < strength) = £ fstress (x)Rstrength (x)

x=0

where the strength and stress are in the positive domain. When stress ~ EDHE(0i, 02, ai) and strength^ EDHE(03,04, a2), the expected reliability is

R = £ V(x + 1; 01, 02, «1) — V(x; 01,02, «2)

x=0 (f2 — f1)a1

' V (x; 0з, 04, «2 )

(04 - 03)k2

Tables 1-4 show the numerical values of R for different values of stress-strength parameters.

Table 1: Values of R for 0i = 0.1, 02 = 0.3, 03 = 0.3, 04 = 0.6 and different values of ai and a2

ai

a1 0.2 0.6 1 1.5 2 2.5

0.2 0.5185 0.6342 0.6962 0.7409 0.7685 0.7874

0.6 0.1707 0.2952 0.3708 0.4319 0.4735 0.5040

1 0.0735 0.1588 0.2166 0.2676 0.3050 0.3338

1.5 0.0336 0.0834 0.1207 0.1564 0.1843 0.2068

2 0.0181 0.0479 0.0718 0.0960 0.1157 0.1321

Table 2: Values of R for 0i = 0.3, 02 = 0.5, 03 = 0.6, 04 = 0.8 and different values of ai and a2

a2

«1 0.2 0.6 1 1.5 2 2.5

0.2 0.6338 0.7240 0.7777 0.8185 0.8441 0.8617

0.6 0.2799 0.4162 0.5027 0.5733 0.6210 0.6557

1 0.1430 0.2657 0.3487 0.4210 0.4730 0.5127

1.5 0.0753 0.1690 0.2373 0.3011 0.3499 0.3890

2 0.0469 0.1169 0.1712 0.2249 0.2680 0.3037

Table 3: Values of R for 0i = 0.5, 02 = 0.8, 03 = 0.5, 04 = 0.8 and different values of ai and a2

a2

«1 0.2 0.6 1 1.5 2 2.5

0.2 0.7372 0.8258 0.8748 0.9092 0.9293 0.9421

0.6 0.4272 0.5967 0.6952 0.7686 0.8138 0.8442

1 0.2720 0.4598 0.5747 0.6649 0.7235 0.7646

1.5 0.1759 0.3550 0.4713 0.5685 0.6352 0.6838

2 0.1277 0.2886 0.3991 0.4962 0..5658 0.6183

Table 4: Values of R for «i=«2 = 0.6, and different values of 0i, 02, 03, 04

02 = 04

01 = 03 0.6 0.7 0.8 1 1.5

0.1 0.4940 0.4974 0.50000 0.5037 0.5091

0.2 0.5437 0.5470 0.54972 0.5540 0.5611

0.3 0.5627 0.5668 0.5704 0.5761 0.5863

0.4 0.5751 0.5803 0.5848 0.5923 0.6058

0.5 0.5850 0.5912 0.5967 0.6059 0.6226

From tables 1-3 it is clear that for fixed values of 01, 02, 03, 04 and a\ reliability increases as a2 tends to infinity. But the reliability decreases with a1 tends to infinity for fixed values of 01, 02, 03, 04 and a2 . Table 4 shows that reliability increases with increasing values of 01, 03 for fixed values of 02, 04, «1 and «2. Also for fixed values of 01, 03, «1 and «2, reliability increases with increasing values of 02 , 04.

2.8. Estimation

In this section we estimate the parameters 01,02 and a of EDHE distribution using the method of maximum likelihood. Let us take a random sampleX1, X2...Xn of size n from EDHE distribution. Then the logarithm of likelihood function is

logL = E log

x=0

V(x + 1; 0i, 02, a) - V(x; 0i, 02, a)

(02 - 01 )a

The maximum likelihood estimators of 01, 02 and a are obtained by solving the equations

dlogL _ dlogL

d0i

0,-

0,

dlogL da

0.

But these equations cannot be solved analytically. So we use Non Linear Maximization (nlm) method for estimating the parameters 01, 02 and a.

3. Simulation Study

In this section, we use Monte-Carlo simulation method to illustrate the performance of the nlm

estimator of the parameters 01 and 02 and a. We generate 5000 random samples of sizes n=20, 30, 75 and 100 from the HE(01,02) distribution for some true values of the parameter set (01,02) = (15,18), (15,21), (16,18) and (16,21). We discretize the generated data and find out 5000 estimates of 01 and 02 and a using (4)for each sample sizes. The estimate of the parameter, average bias and mean square error of the estimate (MSE) are computed and it is given in Table 5 to Table 12.

Table 5: Values of estimates, average bias and average MSE for 01=15 and different values of, 02, a and n=20.

01 02 a 01 Bias(01 ) MSE(01 ) 02 Bias(02) MSE(02) a Bias(a) MSE(a)

15 18 0.4 14.5668 -0.4331 0.1876 17.8307 -0.1693 0.0287 0.3775 -0.0224 0.00005

0.8 14.4120 -0.5879 0.3457 17.7375 -0.2624 0.0689 0.7133 -0.0866 0.0075

15 21 0.4 14.8996 -0.1003 0.0101 20.3668 -0.6331 0.4009 0.3370 -0.0629 0.004

0.8 14.3655 -0.6344 0.4025 20.6269 -0.3730 0.1392 0.6763 -0.1236 0.0153

Table 6: Values of estimates, average bias and average MSE for 01=15 and different values of 02, a and n=30.

01 02 a 01 Bias(01 ) MSE(01 ) 02 Bias(02) MSE(02) a Bias(a) MSE(a)

15 18 0.4 14.5689 -0.4310 0.1858 17.8316 -0.1683 0.0283 0.3807 -0.0192 0.00004

0.8 14.5509 -0.4490 0.2016 17.5699 -0.4300 0.185 0.7356 -0.0643 0.0041

15 21 0.4 14.9042 -0.0957 0.0092 20.6514 -0.3485 0.1215 0.3487 -0.0512 0.0026

0.8 14.5340 -0.4659 0.2171 20.6291 -0.3708 0.1375 0.6883 -0.1116 0.0125

Table 7: Values of estimates, average bias and average MSEfor p\=15 and different values of p2, a and n=75.

$1 $2 a $1 Bias($ ) MSE($ ) $2 Bias($? ) MSE($?) a Bias(a) MSE(a)

15 18 0.4 0.8 14.6128 14.4120 -0.3871 -0.5879 0.1499 0.3457 17.8372 17.7375 -0.1627 -0.2624 0.0265 0.0689 0.4090 0.7133 0.0099 -0.0866 0.00001 0.0075

15 21 0.4 0.8 14.9614 14.6453 -0.0380 -0.3546 0.0014 0.1258 20.8440 20.6317 -0.1550 -0.3683 0.0241 0.1356 0.3502 0.6979 -0.0497 -0.1020 0.0024 0.0104

Table 8: Values of estimates, average bias and average MSEfor p\=15 and different values of p2, a and n=100.

$1 $2 a $1 Bias(^1 ) MSE$ ) $2 Bias(^2) MSE(^2) a Bias(a) MSE(a)

15 18 0.4 0.8 14.6154 14.7504 -0.3845 -0.2494 0.1478 0.0622 17.8376 17.8349 -0.1623 -0.2494 0.0263 0.0273 0.4089 0.7988 0.0089 -0.0011 0.00001 0.0000

15 21 0.4 0.8 15.0249 15.0320 0.0249 0.0320 0.00006 0.0010 20.8483 20.65213 -0.1516 -0.3478 0.0230 0.1210 0.37770 0.7445 -0.0226 -0.0554 0.00005 0.0031

Table 9: Values of estimates, average bias and average MSEfor pi=16 and different values of p2, a and n=20.

$1 $2 a $1 Bias($ ) MSE($ ) $2 Bias($2) MSE($2) a Bias(a) MSE(a)

16 18 0.4 0.8 15.4568 15.6653 -0.5431 -0.3346 0.2950 0.1120 17.8664 17.4611 -0.1335 -0.5388 0.0178 0.2903 0.3409 0.6452 -0.0590 -0.1547 0.0035 0.0240

16 21 0.4 0.8 15.1495 15.7852 -0.8504 -0.2147 0.7233 0.0461 20.4267 20.4893 -0.5732 -0.5101 0.3286 0.2603 0.3003 0.6690 -0.0996 -0.1309 0.0099 0.0172

Table 10: Values of estimates, average bias and average MSEfor p\=16 and different values of p2, a and n=30.

$1 a $1 Bias($ ) MSE($ ) Bias($2 ) MSE($?) a Bias(a) MSE(a)

16 18 0.4 0.8 15.4593 15.8042 -0.5406 -0.1957 0.2923 0.0383 17.8662 17.4674 -0.1337 -0.5325 0.0179 0.2836 0.3452 0.6708 -0.0547 -0.1291 0.0030 0.0167

16 21 0.4 0.8 15.1507 15.8184 -0.8492 -0.1815 0.7212 0.0330 20.4282 20.4918 -0.5717 -0.5081 0.3269 0.2583 0.3008 0.7341 -0.0991 -0.0658 0.0098 0.0043

Table 11: Values of estimates, average bias and average MSEfor pi=16 and different values of p2, a and n=75.

$1 $2 a $1 Bias($1 ) MSE($1 ) $2 Bias($2) MSE($2) a Bias(a) MSE(a)

16 18 0.4 0.8 15.4822 15.8212 -0.5177 -0.1787 0.2681 0.0319 17.8708 17.5036 -0.1291 -0.4963 0.0167 0.2463 0.3791 0.7450 -0.0208 -0.0549 0.00004 0.0030

16 21 0.4 0.8 15.1631 15.9054 -0.8368 -0.0945 0.7002 0.0089 20.4289 20.5039 -0.5710 -0.4961 0.3261 0.2461 0.3171 0.7946 -0.08281 0.0053 0.0069 0.0012

Table 12: Values of estimates, average bias and average MSEfor pi=16 and different values of p2, a and n=100.

$1 $2 a $1 Bias($1 ) MSE($1 ) $2 Bias($2) MSE($2) a Bias(a) MSE(a)

16 18 0.4 0.8 15.4951 15.9119 -0.5048 -0.0880 0.2548 0.0078 17.8731 17.9695 -0.1268 -0.0304 0.0161 0.00009 0.3989 0.7942 -0.0010 -0.0057 0.0000 0.0000

16 21 0.4 0.8 15.1645 15.9066 -0.8354 -0.0933 0.6979 0.0087 20.4290 20.5044 -0.5710 -0.4956 0.3261 0.2456 0.3183 0.8010 -0.0816 0.0010 0.0067 0.0000

From tables 5-12, it is clear that as sample size increases, the average bias and average MSE becomes very small for different choices of the values of the parameters. This indicates the consistency of the estimators.

4. Real Data Analysis

For studying the efficiency of EDHE distribution we consider the data set used by Krishna and Pundir [15] and it represents the total number of carious teeth among the four deciduous molars in a sample of 100 children 10 and 11 years of old. The data are given in Table 13.

Table 13: Observed data

X 0 1 2 3 4

f 64 17 10 6 3

Figure 3 shows the observed data.

Figure 3: Observed data.

We fit the EDHE distribution using the empirical data set and the embeded figure is given in Figure 4.

Figure 4: Embeded figure.

In order to assess the suitability of the proposed model, we use chi-square test of goodness of fit. Also, we compare the EDHE distribution with discrete Lindley (DL) distribution discrete Pareto (DP) distribution and the values of Log-likelihood, AIC, BIC are computed and is shown in Table 14.

Table 14: MLE's, Chi-square value, -Log-likelihood value, AIC values, BIC values and P values for the observed data.

Distribution fitted estimators Chi-square -LL value AIC value BIC value P value

DLD d = 0.275 6.637 113.68 229.36 229.36 0.036

DPD P = 0.1837 3.226 116.83 235.66 235.66 0.199

EDHED 01 = 0.9824779 (p2 = 0.9824794 a = 0.3346 1.2611 111.54 229.08 229.08 0.8679

From Table 14, it is inferred that the EDHE distribution is a better fit than discrete Lindley

and discrete Pareto distributions.

5. Generalizations

5.1. Transmuted exponentiated discrete hypo exponential (TEDHE) distribution

Many transmuted distributions are proposed and studied in literature. For details see Rahman et al. [29] and Dey et al. [6]. In this section we present a generalization of (4) called the transmuted exponentiated discrete hypo exponential distribution. A random variable X is said to have transmuted distribution if its distribution function and probability mass functions are respectively given by

and

F(x) = G(x)[1 + fi - fiG(x)]; №< 1

P(X = x) = g(x)[1 + fi - 2fiG(x)]

(5)

(6)

where G(x), g(x) are the distribution function and probability mass function of the baseline distribution. Also if fi = 0, we will get the baseline distribution. By using equations (5) and (6), the distribution function and probability mass function of the TEDHE distribution is obtained as

F(x; 01,02, a, fi)

V (x; 01,02, a) (02 - 01 )a

1 + fi - fi

V(x; 01,02, a) (02 - 01)a

and

f (x; 01,02, a, fi)

V(x + 1; 01,02, a) - V(x; 01,02, a) (02 - 01 )a

1 + fi - 2fi

V(x; 01,02, a) (02 - 01)a

(7)

(8)

The plot of pmf of TEDHE distribution is given in Figure 5.

Figure 5: Plot of pmf of TEDHE distribution.

The survival function and hazard rate functions are given by the expressions

S(x) = 1 -

V(x; 01,02, a) (02 - 01 )a

1 + fi - fi

V(x; 01,02, a) (02 - 01 )a

and

h(x)

V(x+1;01,02,a)-V(x;01,02,a)

(02-01 )a

1+fi - 2fi(

1 _ V(x;01,02,a)

1 (02-01 )a

1+fi - fii

The hazard plots of TEDHE distribution is given in Figure 6.

Figure 6: Plot of hazard function ofTEDHE distribution.

From Figure 6, it is understood that for different model parameters, the hazard rate function can be decreasing, increasing and increasing-decreasing, which makes the TEDHE distribution more flexible and can model different types of data sets.

5.2. Marshall-Olkin exponentiated discrete hypo exponential (MOEDHE)

distribution

Marshall and Olkin [20] introduced a new method for adding a parameter d(> 0) to the baseline distribution in order to generalize it. Using this method many generalized distributions are proposed and for a detailed review see Gillariose et al. [12]. If F(x) is the survival function of a distribution, then, by Marshall-Olkin method, another survival function G(x) is obtained as

G(x, 9)

9 F(x)

1 - (1 - d)F(x)

=—; -to < X < TO, 9 > 0.

The corresponding distribution function, probability mass function and hazard rate is obtained as

F(x)

G(x, 9)

g(x,9)

h(x)

1 - (1 - 9)F(x) G(x,9) - G(x - 1,9)

9f (x)

[1 - (1 - 9)F(x)][1 - (1 - 9)F(x - 1)]

g_(x) G(x).

where f(x) is the probability mass function corresponding to the distribution function F(x). Using Marshall-Olkin method the survival function of MOEDHE distribution is

G(x, 9)

9(i Yixxj$1j$ial) 9(1 to-^r )

1 - [(1 - 9)(1 - V^-^l)]

The corresponding distribution function, probability mass function and hazard rate are respectively given by

G(x, 9)

g(x,9)

and

h(x)

V(x;fa,4>2A) to

1 - [(1 - 9)(1 - V^-f^)]

9 V(x+l;<p1,<p2,K)-V(x;<p1,<p2,K) 9 (<p2-<h )a

(1 - [(1 - 9)(1 - Vx-ffi)])(1 - [(1 - 9)(1 - V%-1f)22A))])

V(x+1;<ft1,<fe,a)-V (x;<fr,<fe,a) (02-$1)x

(1 - [(1 - 9)(1 - V(X--$A) )])(1 -

The plot of probability mass function of MOEDHE distribution is given in Figure 7.

Figure 7: Plot of pmf of MOEDHE distribution.

The hazard plots are given in Figure 8.

Figure 8: Hazard plots of MOEDHE distribution.

Figure 8 shows different shapes of hazard rate functions and so we can conclude that the MOEDHE distribution is a flexible model in modeling different types of data sets.

6. Summary

Recently, there has been thriving interest in developing new families of distributions by adding one or more additional parameters to the baseline distributions. The existence of various generalization techniques were attracted by many researchers and using one among them we proposed and studied a new distribution called exponentiated discrete hypo exponential distribution. Various distributional and structural properties of this distribution are studied. Also stress-strength analysis is carried out. To evaluate the performance of the nlm estimator, we conducted a simulation study and found that the nlm estimator is consistent. A real data application is carried out and inferred that our proposed distribution is better model than discrete Lindley and discrete Pareto distribution. Two generalizations of the proposed distribution namely transmuted exponentiated discrete hypo exponential distribution and Marshall-Olkin exponentiated discrete hypo exponential distribution are introduced.

References

[1] Ahmad, A. A. and Ghazal, M.G. M. (2020). Exponentiated additive Weibull distribution. Reliability Engineering and System Safety, 193, 106663.

[2] Al-Sulami, D. (2020). Exponentiated Exponential Weibull Distribution. American Journal of Applied Sciences, 17,188-195.

[3] Baharith, L. A and Alamoudi, H. H. (2021). The exponentated Frechet Generator of distributions with applications. Symmetry, 13(4), 572.

[4] Cordeiro, G. M and Castro, M. (2011). A new family of generalized distributions. J.Stat.Comput.Simul., 81, 883-898.

[5] Cordeiro, G. M. ,Ortega, E. M. M and Silva, G. (2012). The Beta extended Weibull family. J.Prob.Stat.Sci., 10,15-40.

[6] Dey, S., Kumar, D.,Anis, M. Z. and Nadarajah, S. (2021). A review of transmuted distributions. Journal of the Indian Society for Probability and Statistics, 22, 47-111.

[7] El-Bassiouny, A. H., EL-Damcese, M., Abdelfattah, M. and Eliwa, M. S. (2017). Exponentiated generalized Weibull-Gompertz distribution with application in survival analysis. Journal of Statistics Applications and Probability, 6, 7-16.

[8] El-Morshedy, M., El-Bassiouny, A. H. and El-Gohary, A. (2017). Exponentiated Inverse flexible Weibull extension distribution. Journal of Statistics Applications and Probability, 6, 169-183.

[9] El-Morshedy, M., Eliwa, M. S., El-Gohary, A and Khalil, A. A. (2020). Bivariate exponentiated discrete Weibull distribution: Statistical Properties,estimation, Simulation and Applications. Mathmatical Sciences , 14(1).

[10] El-Morshedy, M., Eliwa, M. S. and Nagy, H. (2019). A new two-parameter exponentiated discrete Lindley distribution: Properties,estimation and Applications. Journal of Applied Statistics .

[11] Eugene, N., Lee, C and Famoye, F. (2002). Beta-normal distribution and its applications. Communications in Statistics-Theory and Methods, 31, 497-512.

[12] Gillariose, J., Tomy, L., Chesneau, C. and Jose, M. (2020). Recent developments in Marshall-Olkin distributions. Contributions to Mathematics,2, 71-75.

[13] Gomez-Deniz, E. (2010). Another Generalization of the geometric distribution. Test 19, 399-415

[14] Gupta, R. D and kundu, D. (1999). Generalized exponential distribution, Aust.N.Z.J.Stat., 41, 173-188.

[15] Krishna, H and Pundir, P. S. (2009). Discrete Burr and discrete Pareto distributions. Statistical Methodology, 6, 177-188.

[16] Lehmann, E. L. (1953). The power of rank tests. Ann. Math. Statistics, 24, 23-43.

[17] Lemonte, W., Souza, w. B and Cordeiro, G. M. (2013). The exponentiated Kumaraswamy distribution and its log-transform. Brazillian Journal of Probability and Statistics, 27, 31-53.

[18] Liyange, W and Pararai, M. (2014). A generalized power lindley distribution with applications. Asian Journal of Math. Applications and Probability, 18,1-23.

[19] Mahmoudi, E and Zakerzadeh, H. (2010). Generalized Poisson-Lindley distribution. Communications in Statistics-Theory and Methods, 39,1785-1798.

[20] Marshall, A. W. and Olkin, I. (1997). A new method for adding a parameter to a family of distributions with applications to the exponential and Weibull families. Biometrica, 84(3), 641-652.

[21] Mashhadzadeh, Z. H and Mirmostafaee, S. M. T. K. (2020). The exponentiated discrete inverse Rayleigh distribution. Journal of Hyperstructures, 9(1), 54-61.

[22] Merovci, F. and Elbatal, I. (2014). Transmuted Lindley - Geometric distribution and its applications. Journal of Statistics Applications and Probability, 3, 77-91.

[23] Merovci, F. and Sharma, V. K. (2014). The Beta-Lindley distribution: Properties and Applications. Journal of Applied Mathematics.

[24] Nadarajah, S. (2011). The exponentiated exponential distribution:asurvey. AStAAdv. Stat. Anal., 95,219-251.

[25] Nadarajah, S. and Bakar, S. S. A. (2015). An exponentiated geometric distribution. Applied Mathematical Modelling, 1-10.

[26] Nadarajah, S. Bakouch, H. S. and Tahmasbi, R. (2011). A generalized Lindley distribution.

Sankhya B 73 , 331-359.

[27] Nekoukhou, V. and Bidram, H. (2015a). The exponentiated discrete Weibull distribution. SORT 39, 127-146.

[28] Pal, M., Ali, M. M and Woo, J. (2006). Exponentiated Weibull Distribution. Statistica, 66(2), 139-147.

[29] Rahman, M. M, Zahrani, B. L.,Shahbaz, S. H. and Shahbaz, M. Q. (2020). Transmuted probability distributions: A Review. Pakistan Journal of Statistics and Operations Research, 16(1), 83-94.

[30] Ristic, M. M and Balakrishnan, N. (2012). The gamma exponentiated exponential distribution. J.Stat.Comput.Simul., 82, 1191-1206.

[31] Steutel, F. W and Van Harn, K. Infinite divisibility of Probability Distributions on the Real Line. Vol.259, New York, Marcel Dekker, 2004.

[32] Tahir, M. H and Nadarajah, S. (2015). Parameter induction in continuous univariate distributions: Well established G-families. Anais da Academia Brasileira de Cincias , 87(2), 539-568.

[33] Zografos, K and Balakrishnan, N. (2009). On families of Beta and generalized Gamma generated distributions and associated Inference. Statistical Methodology, 344-362.