Hindawi

Journal of Probability and Statistics

Volume 2017, Article ID 2106748, 9 pages
https://doi.org/10.1155/2017/2106748

Research Article
New Generalizations of Exponential
Distribution with Applications

N. A. Rather and T. A. Rather

Department of Mathematics, University of Kashmir, Srinagar 190006, India

Correspondence should be addressed to T. A. Rather; tariqrather@rediffmail.com

Received 5 December 2016; Revised 26 March 2017; Accepted 7 June 2017; Published 6 July 2017

Academic Editor: Chin-Shang Li

Copyright © 2017 N. A. Rather and T. A. Rather. This is an open access article distributed under the Creative Commons Attribution
License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

The main purpose of this paper is to present 𝑘-Generalized Exponential Distribution which among other things includes
Generalized Exponential and Weibull Distributions as special cases. Besides, we also obtain three-parameter extension of
Generalized Exponential Distribution. We shall also discuss moment generating functions (MGFs) of these newly introduced
distributions.

1. Introduction The Euler gamma function Γ(𝛼) is defined by the integral

∞
The gamma family distributions was discussed by Karl
Γ (𝛼) = ∫ 𝑡𝛼−1 𝑒−𝑡 𝑑𝑡, 𝑅 (𝛼) > 0. (1)
Pearson in 1895 as pointed out in Balakrishnan and Basu 0
[1]. However, after a period of 35 years the Exponential
Distribution is a special case of the gamma distribution to A random variable 𝑋 is said to have gamma distribution with
appear on its own. It is also related to Poisson process as parameter 𝛼 > 0, if its p.d.f. is given by
it has been observed that the time between two successive 1 𝛼−1 −𝑥
Poisson events follows the Exponential Distribution. While 𝑓 (𝑥) = 𝑥 𝑒 , 0<𝑥<∞
Γ (𝛼) (2)
discussing the sampling of standard deviation (SD), the
Exponential Distribution was referred to by Kondo [2] as 𝑓 (𝑥) = 0, elsewhere.
Pearson’s Type 𝑋 distribution. Steffensen [3], Teissier [4], and
Weibull [5] proposed the applications of Exponential Dis- Replacing 𝑥 by 𝑥/𝜆, we get the following form of gamma
tribution in actuarial, biological, and engineering problems, distribution with parameters 𝛼, 𝜆 with 𝛼 > 0 and 𝜆 > 0:
respectively.
1
An extension of Exponential Distribution was proposed 𝑓 (𝑥) = 𝑥𝛼−1 𝑒−𝑥/𝜆 ,
by Weibull (1951). The Exponential Distribution is a special 𝜆𝛼 Γ (𝛼) (3)
case wherein the shape parameter equals one. The Weibull
0 < 𝑥 < ∞, 𝛼 > 0, 𝜆 > 0.
distribution has many applications in survival analysis and
reliability engineering; for reference see Lai et al. (2006). The gamma distribution with parameters 𝛼, 𝜆 often arises in
Some other applications in industrial quality control are practices, as the distribution of time one has to wait until a
discussed in Berrettoni (1964). fixed number of events have occurred.
The beta function of two variables 𝑚 and 𝑛 is defined by
2. Some Basic Definitions and 𝐾-Generalized 1
Exponential Distributions 𝐵 (𝑚, 𝑛) = ∫ 𝑥𝑚−1 (1 − 𝑥)𝑛−1 𝑑𝑥,
0 (4)
We begin with some definitions which provide a base for the
definition of 𝐾-Generalized Exponential Distributions. Re (𝑚) > 0, Re (𝑛) > 0.
2 Journal of Probability and Statistics

If we take 𝑥 = 1 − 𝑦, then, 0 ≤ 𝑥 ≤ 1 implies 0 ≤ 𝑦 ≤ 1, and For 𝑅(𝑚) > 0, 𝑅(𝑛) > 0, the 𝑘-beta function of two variables
we get 𝑚 and 𝑛 is defined by
1
𝑚−1 1 1 𝑚/𝑘−1
𝐵 (𝑚, 𝑛) = ∫ (1 − 𝑦) 𝑦𝑛−1 𝑑𝑦 𝐵𝑘 (𝑚, 𝑛) = ∫ 𝑡 (1 − 𝑡)𝑛/𝑘−1 𝑑𝑡. (12)
0 𝑘 0
1 It is implicit in the literature (for reference see [7]) that
𝑚−1
= ∫ 𝑦𝑛−1 (1 − 𝑦) 𝑑𝑦 (5)
0 Γ𝑘 (𝑚) Γ𝑘 (𝑛)
𝐵𝑘 (𝑚, 𝑛) = . (13)
1 Γ𝑘 (𝑚 + 𝑛)
= ∫ 𝑥𝑛−1 (1 − 𝑥)𝑚−1 𝑑𝑥 = 𝐵 (𝑛, 𝑚) .
0 For the sake of completeness, we present a very simple proof
In the literature (for reference see [6–8]), it is known that of the relation (13).
We have
Γ (𝑚) Γ (𝑛)
𝐵 (𝑛, 𝑚) = . (6) 1 1 𝑚/𝑘−1
Γ (𝑚 + 𝑛) 𝐵𝑘 (𝑚, 𝑛) = ∫ 𝑥 (1 − 𝑥)𝑛/𝑘−1 𝑑𝑥. (14)
𝑘 0
In (5), if we take
Put
1
𝑥= (7) 𝑥 = cos2 𝜃; (15)
𝑧+1
then 0 ≤ 𝑥 ≤ 1 implies 0 ≤ 𝑧 < ∞, and we get then
𝐵 (𝑛, 𝑚) 𝑑𝑥 = −2 cos 𝜃 sin 𝜃𝑑𝜃, (16)
0
1 𝑛−1 𝑧 𝑚−1 1 and we get
=∫ ( ) ( ) (− ) 𝑑𝑧
∞ 𝑧+1 𝑧+1 (𝑧 + 1)2 (8)
2 𝜋/2
𝑚−1
𝐵𝑘 (𝑚, 𝑛) = ∫ (cos 𝜃)2𝑚/𝑘−1 (sin 𝜃)2𝑛/𝑘−1 𝑑𝜃. (17)
∞
𝑧 𝑘 0
=∫ 𝑑𝑧.
0 (𝑧 + 1)𝑚+𝑛 Now
∞
A continuous random variable 𝑋 is said to have beta distri- 𝑘
Γ𝑘 (𝑚) = ∫ 𝑡𝑚−1 𝑒−𝑡 /𝑘
𝑑𝑡. (18)
bution with parameters 𝑚 and 𝑛, if its p.d.f. is given by 0

1 Put
𝑓 (𝑥) = 𝑥𝑚−1 (1 − 𝑥)𝑛−1 , 0 ≤ 𝑥 ≤ 1
𝐵 (𝑚, 𝑛) (9) 𝑡𝑘
= 𝑥2 (19)
𝑓 (𝑥) = 0, elsewhere. 𝑘
or
This distribution is known as beta distribution of 1st kind (for
1/𝑘
reference see [7]). 𝑡 = (𝑘𝑥2 ) , (20)
The beta distribution has an application to model a random
phenomenon whose set of possible values is a finite interval so that
[𝑎, 𝑏], which by letting 𝑎 denote the origin and taking (𝑏 − 𝑎) 2 1/𝑘 2/𝑘−1
as a unit measurement can be transformed into the interval 𝑑𝑡 = 𝑘 𝑥 𝑑𝑥 = 2𝑘1/𝑘−1 𝑥2/𝑘−1 𝑑𝑥; (21)
[0, 1]. 𝑘
A continuous random variable 𝑋 is said to have beta distribu- we get
tion of 2nd kind with parameters 𝑚 and 𝑛, if its p.d.f. is given ∞ 2
by Γ𝑘 (𝑚) = 2𝑘1/𝑘−1 ∫ 𝑘(𝑚−1)/𝑘 𝑥(2/𝑘)(𝑚−1) 𝑒−𝑥 𝑥2/𝑘−1 𝑑𝑥
0
1 𝑧𝑚−1 ∞
(22)
𝑓 (𝑥) = , 0 ≤ 𝑧 < ∞, 𝑚, 𝑛 > 0, (𝑚/𝑘)−1 2𝑚/𝑘−1 −𝑥2
𝐵 (𝑚, 𝑛) (1 + 𝑧)𝑚+𝑛 (10) = 2𝑘 ∫ 𝑥 𝑒 𝑑𝑥.
0
𝑓 (𝑥) = 0, elsewhere. Since the integrals involved are convergent, we have
More recently, Rahman et al. [7] (for more details see [8– ∞ 2

16]) have defined 𝑘-gamma and 𝑘-beta distributions and their Γ𝑘 (𝑚) Γ𝑘 (𝑛) = (2𝑘𝑚/𝑘−1 ∫ 𝑥2𝑚/𝑘−1 𝑒−𝑥 𝑑𝑥)
0
MGFs as follows.
∞
For 𝑘 > 0 and 𝑧 ∈ L, the 𝑘-gamma function is defined by 2
⋅ (2𝑘𝑛/𝑘−1 ∫ 𝑦2𝑛/𝑘−1 𝑒−𝑦 𝑑𝑦) (23)
the integral 0
∞ 𝑘 ∞ ∞ 2 2
Γ𝑘 (𝑧) = ∫ 𝑡𝑧−1 𝑒−𝑡 /𝑘
𝑑𝑡. (11) = 4𝑘(𝑚+𝑛)/𝑘−2 ∫ ∫ 𝑥2𝑚/𝑘−1 𝑦2𝑛/𝑘−1 𝑒−(𝑥 +𝑦 ) 𝑑𝑥 𝑑𝑦.
0 0 0
Journal of Probability and Statistics 3

Put Putting 𝑚 = 𝑘/2 and 𝑛 = 𝑘/2 in (17), we get

𝑥 = 𝑟 cos 𝜃,
𝑘 𝑘 2 𝜋/2 2 2 𝜋
(24) 𝐵𝑘 ( , ) = ∫ 𝑑𝜃 = [𝜃]𝜋/2 = ( )( )
𝑦 = 𝑟 sin 𝜃, 2 2 𝑘 0 𝑘 0 𝑘 2
(33)
𝜋
so that = ,
𝑘
𝑑𝑥𝑑𝑦 = 𝑟𝑑𝑟𝑑𝜃; (25)
which establishes (ii).
we get Putting the value of 𝐵𝑘 (𝑘/2, 𝑘/2) in (32), we get
∞ 𝜋/2
𝑘 𝜋
Γ𝑘 (𝑚) Γ𝑘 (𝑛) = 4𝑘(𝑚+𝑛)/𝑘−2 ∫ ∫ (cos 𝜃)2𝑚/𝑘−1 Γ𝑘 ( ) = √ , (34)
0 0 2 𝑘
2 2
⋅ (sin 𝜃)2𝑛/𝑘−1 𝑒−𝑟 𝑟2𝑚/𝑘+2𝑛/𝑘−1 𝑑𝑟 𝑑𝜃 = { which proves (iii).
𝑘
(26) A random variable 𝑋 of continuous type is said to have
𝜋/2
⋅∫ 2𝑚/𝑘−1
(cos 𝜃) (sin 𝜃) 2𝑛/𝑘−1
𝑑𝜃} Weibull distribution if its probability density function is given
0 by
∞
⋅ {2𝑘(𝑚+𝑛)/𝑘−1 ∫ 𝑟2(𝑚+𝑛)/𝑘−1 𝑒−𝑟 𝑑𝑟} .
2
𝛽 𝑥 − ] 𝛽−1 −((𝑥−])/𝛼)𝛽
𝑓 (𝑥, 𝛼, 𝛽, ]) = ( ) 𝑒 , 𝑥>]
0 𝛼 𝛼 (35)
Using (17) and (22), we get 𝑓 (𝑥, 𝛼, 𝛽, ]) = 0, elsewhere, 𝑥 ≤ ].
Γ𝑘 (𝑚) Γ𝑘 (𝑛) = 𝐵𝑘 (𝑚, 𝑛) Γ𝑘 (𝑚 + 𝑛) , (27) Weibull distribution is widely used in engineering prac-
which gives tice due to its versatility. It was originally proposed for
the interpretation of fatigue data but now it is also used
Γ𝑘 (𝑚) Γ𝑘 (𝑛) for many other problems in engineering. In particular, in
𝐵𝑘 (𝑚, 𝑛) = . (28) the field of life phenomenon, it is used as the distribution
Γ𝑘 (𝑚 + 𝑛)
of lifetime of some object, particularly when the “weakest
This completes the proof of the relation (13). link” model is appropriate for the model; that is, consider
an object consisting of many parts and suppose that the
Corollary 1. One has the following: object experiences death (failure) when any of its parts fail;
(i) Γ𝑘 (𝑘) = 1. it has been shown [6] (both oretically and empirically) under
these conditions that Weibull distribution provides a close
(ii) 𝐵𝑘 (𝑘/2, 𝑘/2) = 𝜋/𝑘. approximation to the distribution of the lifetime of the item.
(iii) Γ𝑘 (𝑘/2) = √𝜋/𝑘. The Gamma and Weibull distributions are commonly
used for analyzing any lifetime data or skewed data. Both
Proof. Put 𝑚 = 𝑘 in (22); we get distributions have nice physical interpretation and several
∞
desirable properties. Unfortunately both distributions have
2 2 ∞
Γ𝑘 (𝑘) = 2 ∫ 𝑥𝑒−𝑥 𝑑𝑥 = − [𝑒−𝑥 ] = 1. (29) drawbacks, one major disadvantage of the gamma distribu-
0 0 tion is that the distribution function or the survival function
From (13), we have can not be computed easily if the shape parameter is not an
integer. By using mathematical tables or computer software
Γ𝑘 (𝑚) Γ𝑘 (𝑛) one obtains the distribution function, the survival function,
𝐵𝑘 (𝑚, 𝑛) = . (30) or hazard function. This makes the gamma distribution
Γ𝑘 (𝑚 + 𝑛)
unpopular as compared to Weibull distribution whose distri-
We take 𝑚 = 𝑛 = 𝑘/2; we get bution function, hazard function, or survival function is easy
to compute. It is well known that even though the Weibull
𝑘 𝑘 Γ (𝑘/2) Γ𝑘 (𝑘/2) 𝑘 𝑘
𝐵𝑘 ( , ) = 𝑘 = Γ𝑘 ( ) Γ𝑘 ( ) distribution has convenient representation of distribution
2 2 Γ𝑘 (𝑘) 2 2 function, the distribution of the sum of independent and
(31) identically distributed (i.i.d) Weibull random variables is not
𝑘 2
= [Γ𝑘 ( )] . simple to obtain. Therefore, the distribution of the mean
2 of random sample from Weibull distribution is not easy to
This gives compute whereas the distribution of sum of independent
and identically distributed (i.i.d) gamma random variables is
well known. For more details see Mudholkar, Srivastava and
𝑘 𝑘 𝑘
[Γ𝑘 ( )] = √ 𝐵𝑘 ( , ). (32) Freimer (1995), Mudholkar and Srivastava [17], Gupta and
2 2 2 Kundu [18], and Gupta et al. [19].
4 Journal of Probability and Statistics

Recently R. D. Gupta and D. Kundu have introduced it 𝑘-Generalized Exponential Distribution (𝑘-GED). In fact,
three-parameter Exponential Distribution (location, scale, we prove the following result, which included Generalized
and shape) and studied the theoretical properties of this fam- Exponential Distribution as a special case.
ily and compared them with respective good studies proper-
ties of Gamma and Weibull distributions. The increasing and Theorem 2. Let 𝑋 be a random variable of continuous type
decreasing hazard rate of the Generalized Exponential Distri- and let 𝛼 > 0, 𝛽 > 0, and 𝑘 > 0 be the parameters; then the
bution (GED) depends on the shape parameter. Generalized function
Exponential Distribution (GED) has several properties that
𝑘 𝛼−1 𝑘
are quite similar to gamma distribution but it has distribution 𝑓 (𝑥, 𝛼, 𝛽, 𝑘) = 𝛼𝛽 (1 − 𝑒−𝛽𝑥 /𝑘
) 𝑥𝑘−1 𝑒−𝛽𝑥 /𝑘
,
function similar to that of the Weibull distribution which can
be computed easily. Since the Generalized Exponential family 𝑥 > 0 (40)
has the likelihood ratio ordering on the shape parameter, one
can construct a uniformly most powerful test for testing one 𝑓 (𝑥, 𝛼, 𝛽, 𝑘) = 0, elsewhere
sided hypothesis on the shape parameter when the scale and
location parameters are known. is the p.d.f. of random variable 𝑋 of continuous type.

3. Generalized Exponential Distribution (GED) Remark 3. If we take 𝑘 = 1, it reduces to Generalized

Exponential Distribution.
A continuous random variable 𝑋 whose p.d.f. is given by
Proof of Theorem 2. Clearly
−𝛽𝑥
{𝛽𝑒 if 𝑥 ≥ 0
𝑓 (𝑥) = { 𝛽 > 0, (36) 𝑓 (𝑥, 𝛼, 𝛽, 𝑘) ≥ 0 ∀𝑥 > 0, 𝛼 > 0, 𝛽 > 0, 𝑘 > 0. (41)
0 if 𝑥 < 0,
{
Now
is said to have an Exponential Distribution.
∞
The Generalized Exponential Distribution (GED) intro-
∫ 𝑓 (𝑥, 𝛼, 𝛽, 𝑘) 𝑑𝑥
duced by Mubeen et al. [15] has p.d.f. 0
∞ 𝛼−1
−𝛽𝑥 𝛼−1 −𝛽𝑥 𝑘 𝑘
𝑓 (𝑥, 𝛼, 𝛽) = 𝛼𝛽 (1 − 𝑒 ) 𝑒 , = ∫ 𝛼𝛽 (1 − 𝑒−𝛽𝑥 /𝑘
) 𝑥𝑘−1 𝑒−𝛽𝑥 /𝑘
𝑑𝑥
0
𝑥 > 0 for 𝛼, 𝛽 > 0 (37) ∞ 𝑘 𝛼−1 𝑘
= 𝛼 ∫ (1 − 𝑒−𝛽𝑥 /𝑘
) (𝛽𝑥𝑘−1 𝑒−𝛽𝑥 /𝑘
) 𝑑𝑥 (42)
𝑓 (𝑥, 𝛼, 𝛽) = 0, otherwise. 0

𝑘 𝛼 ∞
−𝛽𝑥 /𝑘
The MGF of (37) is given by [ (1 − 𝑒 ) ] 𝑘 𝛼 ∞
= 𝛼[ ] = [(1 − 𝑒−𝛽𝑥 /𝑘
) ]
∞ 𝛼−1 −𝛽𝑥
𝛼 0
𝑀𝑋 (𝑡) = 𝐸 (𝑒𝑡𝑋 ) = ∫ 𝑒𝑡𝑥 𝛼𝛽 (1 − 𝑒−𝛽𝑥 ) 𝑒 𝑑𝑥 [ ]0
0
(38) = 1 − 0 = 1.
𝑡
= 𝛼𝐵 (𝛼, 1 − ),
𝛽 Hence 𝑓(𝑥, 𝛼, 𝛽, 𝑘) is a p.d.f. of random variable 𝑋 of
continuous type.
1
where 𝐵(𝑚, 𝑛) = ∫0 𝑦𝑛−1 (1 − 𝑦)𝑚−1 𝑑𝑦.
Replacing 𝛽 by 1/𝜆 and 𝑥 by (𝑥 − 𝜇) in (37), we get 5. The Moment Generating Function (MGF)
the following form of Generalized Exponential Distribution
(GED):
of Theorem 2
𝛼 𝛼−1 In this section, we derive MGF of the random variable 𝑋
𝑓 (𝑥, 𝛼, 𝜆, 𝜇) = (1 − 𝑒−(𝑥−𝜇)/𝜆 ) 𝑒−(𝑥−𝜇)/𝜆 , having k-Generalized Exponential Distribution in terms of
𝜆 (39) new parameter 𝑘 > 0; we have
𝑥 > 𝜇, 𝛼 > 0, 𝜆 > 0.
𝑘 ∞ 𝑘
𝑀𝑘 (𝑡) = 𝐸 (𝑒𝑡𝑋 ) = ∫ 𝑒𝑡𝑥 𝑓 (𝑥, 𝛼, 𝛽, 𝑘) 𝑑𝑥
4. Extensions of Generalized Exponential 0
(43)
Distribution (GED) ∞
𝑡𝑥𝑘 −𝛽𝑥
𝑘 /𝑘 𝛼−1
𝑘−1 −𝛽𝑥𝑘 /𝑘
= 𝑘𝛼𝛽 ∫ 𝑒 (1 − 𝑒 ) (𝑥 𝑒 ) 𝑑𝑥.
The main aim of this paper is to present interesting exten- 0
sions of Generalized Exponential Distribution (GED) in
Put
various ways and to study their moment generating functions
(MGFs). We shall first define Generalized Exponential Dis- 𝑘

tribution (GED) in terms of a new parameter 𝑘 > 0 and call 𝑒−𝛽𝑥 /𝑘

= 𝑦; (44)
Journal of Probability and Statistics 5

𝑘 𝛼 ∞
then −𝛽𝑥
[ (1 − 𝑒 ) ]
−𝛽𝑥𝑘
𝛼 ∞
= 𝛼[ ] = [(1 − 𝑒 ) ]
𝑒−𝛽𝑥
𝑘
/𝑘
(−𝛽𝑥𝑘−1 ) 𝑑𝑥 = 𝑑𝑦, 𝛼 0
(45) [ ]0
𝑥𝑘 −𝑘/𝛽
𝑒 =𝑦 = 1 − 0 = 1.
(50)
so that
𝑘 Hence 𝑓(𝑥, 𝛼, 𝛽, 𝑘) is a p.d.f. of random variable 𝑋 of
𝑒𝑡𝑥 = 𝑦−𝑡𝑘/𝛽 . (46) continuous type.

Therefore, Remark 6 (Weibull distribution is a special case of Theo-

rem 4). To see this, we take 𝛼 = 1, 𝛽 = 1 in Theorem 4; it
1
𝛼−1 follows that
𝑀𝑘 (𝑡) = 𝛼 ∫ 𝑦−𝑡𝑘/𝛽 (1 − 𝑦) 𝑑𝑦
0
𝑘 𝑘 𝑑
1
(1−𝑡𝑘/𝛽)−1 𝛼−1 𝑓 (𝑥, 𝑘) = 𝑘𝑥𝑘−1 𝑒−𝑥 = 𝑒−𝑥 (𝑥𝑘 ) , 𝑥 > 0
= 𝛼∫ 𝑦 (1 − 𝑦) 𝑑𝑦 𝑑𝑥 (51)
0
𝑓 (𝑥, 𝑘) = 0, 𝑥 ≤ 0.
1
𝛼−1
= 𝛼 ∫ (1 − 𝑦) 𝑦(1−𝑡𝑘/𝛽)−1 𝑑𝑦 (47)
0 Replacing 𝑥 by (𝑥 − 𝜇)/𝛼, we get
𝑡𝑘 𝛼Γ (𝛼) Γ (1 − (𝑡𝑘/𝛽))
= 𝛼𝐵 (𝛼, 1 − ( )) = 𝑘 𝑑 𝑥−] 𝑘
𝛽 Γ (𝛼 + 1 − (𝑡𝑘/𝛽)) 𝑓 (𝑥, 𝛼, 𝑘, ]) = 𝑒−((𝑥−])/𝛼) ( ) ,
𝑑𝑥 𝛼
Γ (𝛼 + 1) Γ (1 − (𝑡𝑘/𝛽)) 𝑥−]
= . ( ) > 0 (52)
Γ (𝛼 + 1 − (𝑡𝑘/𝛽)) 𝛼
𝑥−]
Our next theorem is also a generalization of Exponential 𝑓 (𝑥, 𝛼, 𝑘, ]) = 0, ( )≤0
Distribution in terms of new variable 𝑘, which includes 𝛼
Weibull distribution as a special case. or
Theorem 4. Let 𝑋 be a random variable of continuous type 𝑘 𝑥 − ] 𝑘−1 −((𝑥−])/𝛼) 𝑘
and let 𝛼 > 0, 𝛽 > 0, and 𝑘 > 0 be the parameters; then the 𝑓 (𝑥, 𝛼, 𝑘, ]) = ( ) 𝑒 , 𝑥 > ],
function 𝛼 𝛼 (53)
𝑓 (𝑥, 𝛼, 𝑘, ]) = 0, 𝑥 ≤ ],
𝑘 𝛼−1 𝑘
𝑓 (𝑥, 𝛼, 𝛽, 𝑘) = 𝑘𝛼𝛽 (1 − 𝑒−𝛽𝑥 ) 𝑥𝑘−1 𝑒−𝛽𝑥 ,
is a p.d.f. of 𝑋, which is clearly the density of Weibull
0 < 𝑥 < ∞ (48) Distribution.

𝑓 (𝑥, 𝛼, 𝛽, 𝑘) = 0, elsewhere,
6. The Moment Generating Function (MGF)
is the p.d.f. of random variable 𝑋 of continuous type. of Theorem 4
Remark 5. For 𝑘 = 1, 𝐾-Generalized Exponential Theorem 4 In this section, we derive MGF of the random variable 𝑋
reduces to Classical Exponential Distribution. having k-Generalized Exponential Distribution in terms of a
new parameter 𝑘 > 0; we have
Proof of Theorem 4. Clearly
𝑘 ∞ 𝑘
𝑀𝑘 (𝑡) = 𝐸 (𝑒𝑡𝑋 ) = ∫ 𝑒𝑡𝑥 𝑓 (𝑥, 𝛼, 𝛽, 𝑘) 𝑑𝑥
𝑓 (𝑥, 𝛼, 𝛽, 𝑘) ≥ 0 ∀𝑥 > 0, 𝛼 > 0, 𝛽 > 0, 𝑘 > 0. (49) 0
(54)
∞ 𝛼−1
𝑡𝑥 𝑘 −𝛽𝑥𝑘 𝑘−1 −𝛽𝑥𝑘
Now = 𝑘𝛼𝛽 ∫ 𝑒 (1 − 𝑒 ) 𝑥 𝑒 𝑑𝑥.
0
∞
∫ 𝑓 (𝑥, 𝛼, 𝛽, 𝑘) 𝑑𝑥 𝑘
0 Put 𝑒−𝛽𝑥 = 𝑦; then
∞ 𝑘 𝛼−1 𝑘
= 𝑘𝛼𝛽 ∫ (1 − 𝑒−𝛽𝑥 ) 𝑥𝑘−1 𝑒−𝛽𝑥 𝑑𝑥 𝑘
0 𝑒𝑥 = 𝑦−1/𝛽 ,
∞ 𝛼−1
(55)
−𝛽𝑥𝑘 𝑘−1 −𝛽𝑥𝑘 −𝛽𝑥𝑘 𝑘−1
= 𝛼 ∫ (1 − 𝑒 ) (𝑘𝛽𝑥 )𝑒 𝑑𝑥 𝑒 (−𝑘𝛽𝑥 ) 𝑑𝑥 = 𝑑𝑦.
0
6 Journal of Probability and Statistics

Therefore, If we replace 𝛽 by 1/𝜆 and 𝑥 by (𝑥−𝜇) in Theorem 7, we get

the following form of Generalized Exponential Distribution
1 (GED):
𝛼−1
𝑀𝑘 (𝑡) = 𝛼 ∫ 𝑦−𝑡/𝛽 (1 − 𝑦) 𝑑𝑦
0 𝑓 (𝑥, 𝛼, 𝜆, 𝜇, 𝛿)
1
𝛼−1 𝛼𝛿 −(𝑥−𝜇)/𝜆 𝛼−1 −(𝑥−𝜇)/𝜆
= 𝛼 ∫ 𝑦(1−𝑡/𝛽)−1 (1 − 𝑦) 𝑑𝑦 = 𝛼 (1 − 𝛿𝑒 ) 𝑒 ,
0 𝜆 (1 − (1 − 𝛿) ) (59)
1
𝛼−1
𝑥 > 𝜇, 𝛼 > 0, 𝜆 > 0, 𝜇 > 0, 0 < 𝛿 ≤ 1.
= 𝛼 ∫ (1 − 𝑦) 𝑦(1−𝑡/𝛽)−1 𝑑𝑦 (56)
0 𝑓 (𝑥, 𝛼, 𝜆, 𝜇, 𝛿) = 0, otherwise.

𝑡 𝛼Γ (𝛼) Γ (1 − (𝑡/𝛽)) Remark 8. For 𝛿 = 1, Theorem 7 reduced to Generalized

= 𝛼𝐵 (𝛼, 1 − ( )) =
𝛽 Γ (𝛼 + 1 − (𝑡/𝛽)) Exponential Distribution (GED).

Remark 9. Taking 𝛿 = 1 in (59), we get relation (39).

Γ (𝛼 + 1) Γ (1 − (𝑡/𝛽))
= .
Γ (𝛼 + 1 − (𝑡/𝛽))
7. The Moment Generating Function (MGF)
We also present the following three-parameter extension of Theorem 7
of Generalized Exponential Distribution (GED). In fact, we
∞
prove the following. 𝑀𝑋 (𝑡) = 𝐸 (𝑒𝑡𝑋 ) = ∫ 𝑒𝑡𝑥 𝑓 (𝑥, 𝛼, 𝛽, 𝛿) 𝑑𝑥
0
Theorem 7. Let 𝑋 be a continuous random variable; then the ∞
function 𝛼𝛽𝛿 𝑡𝑥 −𝛽𝑥 𝛼−1 −𝛽𝑥
= 𝛼 ∫ 𝑒 (1 − 𝛿𝑒 ) 𝑒 𝑑𝑥 (60)
(1 − (1 − 𝛿) ) 0
𝛼𝛽𝛿 −𝛽𝑥 𝛼−1 −𝛽𝑥 𝛼𝛿 ∞ 𝛼−1
𝑓 (𝑥, 𝛼, 𝛽, 𝛿) = 𝛼 (1 − 𝛿𝑒 ) 𝑒 , = ∫ (1 − 𝛿𝑒−𝛽𝑥 ) 𝑒𝑡𝑥 (𝛽𝑒−𝛽𝑥 ) 𝑑𝑥,
1 − (1 − 𝛿) 𝛼
(1 − (1 − 𝛿) ) 0
(57)
𝑥 > 0, 𝛼 > 0, 𝛽 > 0, 0 < 𝛿 ≤ 1, Put

𝑒−𝛽𝑥 = 𝑦, (61)
𝑓 (𝑥, 𝛼, 𝛽, 𝛿) = 0, otherwise
so that
is the p.d.f. of random variable 𝑋 of continuous type.
−𝛽𝑒−𝛽𝑥 𝑑𝑥 = 𝑑𝑦,
Proof of Theorem 7. Clearly 𝑓(𝑥, 𝛼, 𝛽, 𝛿) ≥ 0 for all 𝑥 > 0, 𝛼 > (62)
0, 𝛽 > 0, 𝛿 > 0. 𝑒𝑡𝑥 = 𝑦−𝑡/𝛽 ;
Now we get
∞ 1
∫ 𝑓 (𝑥, 𝛼, 𝛽, 𝛿) 𝑑𝑥 𝛼𝛿 𝛼−1 −𝑡/𝛽
𝑀 (𝑡) = 𝛼 ∫ (1 − 𝛿𝑦) 𝑦 𝑑𝑦,
0 (1 − (1 − 𝛿) ) 0
𝛼𝛽𝛿 ∞
−𝛽𝑥 𝛼−1 −𝛽𝑥 0 < 𝛿 ≤ 1. (63)
= 𝛼 ∫ (1 − 𝛿𝑒 ) 𝑒 𝑑𝑥
1 − (1 − 𝛿) 0
𝛼 𝑡
𝑀 (𝑡) = 𝛼 𝐵𝛿 (𝛼, 1 − ) ,
𝛼 ∞
−𝛽𝑥 𝛼−1
(1 − (1 − 𝛿) ) 𝛽
= 𝛼 ∫ (1 − 𝛿𝑒 ) (𝛽𝛿𝑒−𝛽𝑥 ) 𝑑𝑥
1 − (1 − 𝛿) 0 (58) where
∞ 1
−𝛽𝑥 𝛼 𝑚−1
𝛼 (1 − 𝛿𝑒 ) 𝐵𝛿 = 𝐵𝛿 (𝑚, 𝑛) = ∫ (1 − 𝛿𝑦) 𝑦𝑛−1 𝑑𝑦,
=( 𝛼)
[ ] 0 (64)
1 − (1 − 𝛿) 𝛼
[ ]0 𝑚 > 0, 𝑛 > 0, 0 < 𝛿 ≤ 1.
(1 − (1 − 𝛿)𝛼 ) Remark 10. For 𝛿 = 1, we have
= = 1.
(1 − (1 − 𝛿)𝛼 ) 1
𝑚−1
𝐵1 (𝑚, 𝑛) = ∫ (1 − 𝑦) 𝑦𝑛−1 𝑑𝑦 = 𝐵 (𝑚, 𝑛) . (65)
0
This shows that𝑓(𝑥) is a p.d.f. of the random variable 𝑋. This
proves Theorem 7. And (63) reduces to Remark 8.
Journal of Probability and Statistics 7

1
Remark 11. If in 𝐵𝛿 (𝑚, 𝑛) = ∫0 (1 − 𝛿𝑦)𝑚−1 𝑦𝑛−1 𝑑𝑦, we put 𝑥 = Remark 14. Weibull Distribution is the limiting case of
𝛿𝑦; Theorem 13. To see this, we let 𝛿 → 0 in Theorem 13 and
then note that
1 𝑘𝛿
𝑑𝑦 = 𝑑𝑥 (66) lim = 1, (73)
𝛿 𝛿→0 1 − (1 − 𝛿)𝑘
and we get
so that
𝛿
𝑥 𝑛−1 1
𝐵𝛿 (𝑚, 𝑛) = ∫ (1 − 𝑥)𝑚−1 ( ) 𝑑𝑥 𝛽 𝛽 𝑑
0 𝛿 𝛿 𝑓 (𝑥, 𝛽) = 𝛽𝑒−𝑥 𝑥𝛽−1 = 𝑒−𝑥 (𝑥𝛽 ) , 𝑥 > 0,
(67) 𝑑𝑥 (74)
1 𝛿
= 𝑛 ∫ (1 − 𝑥)𝑚−1 𝑥𝑛−1 𝑑𝑥, 0 ≤ 𝛿 ≤ 1. 𝑓 (𝑥, 𝛽) = 0, 𝑥 ≤ 0,
𝛿 0
Remark 12. Letting 𝛿 → 0 in (57) and noting that is the p.d.f. of random variable 𝑋 of continuous type.
Replacing 𝑥 by (𝑥 − ])/𝛼, it follows that
𝛼𝛿
𝐿𝑡 ⋅ = 1, (68)
𝛿→0 (1 − (1 − 𝛿)𝛼 ) 𝛽 𝑑 𝑥−] 𝛽
𝑓 (𝑥, 𝛼, 𝛽, ]) = 𝛽𝑒−((𝑥−])/𝛼) ( ) ,
we get 𝑑𝑥 𝛼
𝑥−]
𝑓 (𝑥, 𝛽) = 𝛽𝑒−𝛽𝑥 , if 𝑥 ≥ 0, 𝛽 > 0, (69) > 0 (75)
𝛼
which is the p.d.f. of the Exponential Distribution.
𝑥−]
𝑓 (𝑥, 𝛼, 𝛽, ]) = 0, ≤ 0.
Finally we present the following more general interesting 𝛼
result which among other things includes Weibull distribu- Equivalently
tion as a limiting case.
𝛽 𝑥 − ] 𝛽−1 −((𝑥−])/𝛼)𝛽
Theorem 13. Let 𝑋 be a random variable of continuous type. If 𝑓 (𝑥, 𝛼, 𝛽, ]) = ( ) 𝑒 , 𝑥 > ],
𝛿 > 0, 𝛽 > 0, and 𝑘 > 0 are the parameters, then the function 𝛼 𝛼 (76)
𝑘𝛿𝛽 𝛽 𝑘−1 𝛽
𝑓 (𝑥, 𝛼, 𝛽, ]) = 0, 𝑥 ≤ ],
𝑓 (𝑥, 𝛿, 𝛽, 𝑘) = (1 − 𝛿𝑒−𝑥 ) 𝑥𝛽−1 𝑒−𝑥 ,
1 − (1 − 𝛿)𝑘 is a p.d.f. of 𝑋, which is clearly density of Weibull Distribu-
(70) tion.
𝑥 > 0, 0 < 𝛿 < 1,

𝑓 (𝑥, 𝛿, 𝛽, 𝑘) = 0, elsewhere, 𝑥 ≤ 0, 8. Applications

is the p.d.f. of the random variable X. The applications of Exponential Distribution have been
widespread, which include models to determine bout cri-
Proof of Theorem 13. Clearly
teria for analysis of animal behaviour [20]; design rainfall
𝑓 (𝑥, 𝛿, 𝛽, 𝑘) ≥ 0 ∀𝑥 > 0, 𝛿 > 0, 𝛽 > 0, 𝑘 > 0. (71) estimation in the Coast of Chiapas [21]; analysis of Los
Angeles rainfall data [22]; software reliability growth models
Now for vital quality metrics [23]; models for episode peak and
∞
duration for ecohydroclimatic applications [24]; estimating
∫ 𝑓 (𝑥, 𝛿, 𝛽, 𝑘) 𝑑𝑥
0 mean life of power system equipment with limited end-of-
∞ 𝑘−1
life failure data [25]; and cure rate modeling (Kannan et al.
𝛽 𝛽
= 𝑘𝛿𝛽 ∫ (1 − 𝛿𝑒−𝑥 ) 𝑥𝛽−1 𝑒−𝑥 𝑑𝑥 (2010). In related work, the closeness of the exponentiated
0 Exponential Distribution with the Weibull, gamma, and log-
𝑘 ∞ 𝛽 𝑘−1 𝛽 normal distributions is studied in Gupta and Kundu (2003a),
= 𝑘
∫ (1 − 𝛿𝑒−𝑥 ) (𝛽𝛿𝑒−𝑥 𝑥𝛽−1 ) 𝑑𝑥 (2003b), (2004), (2006)) and [26]. Some generalizations of
1 − (1 − 𝛿) 0
the exponentiated Exponential Distribution are discussed in
(72)
𝑘 ∞ Nadarajah and Kotz [27].
−𝑥𝛽
𝑘 [ (1 − 𝛿𝑒 ) ] The Exponential Distribution is often used to model
= [ 𝑘
] the reliability of electronic systems, which do not typically
1 − (1 − 𝛿) 𝑘
[ ]0 experience wear-out type failures. The distribution is called
“memoryless,” meaning that the calculated reliability for,
𝑘 (1 − (1 − 𝛿)𝑘 ) say, a 10-hour mission is the same for a subsequent 10-hour
= ( ) = 1. mission, given that the system is working properly at the start
1 − (1 − 𝛿)𝑘 𝑘
of each mission. Given a hazard (failure) rate, 𝜆, or mean
This shows that 𝑓(𝑥, 𝛼, 𝛽, 𝑘) is the p.d.f. of random variable 𝑋 time between failure (MTBF = 1/𝜆), the reliability can be
of continuous type. This proves Theorem 13. determined at a specific point in time (𝑡). The distribution
8 Journal of Probability and Statistics

function, reliability function, and hazard rate of Theorem 2
are given by
𝑥
𝐹 (𝑥, 𝛼, 𝛽, 𝑘) = 𝑃 (𝑋 ≤ 𝑥) = ∫ 𝑓 (𝑥, 𝛼, 𝛽, 𝑥) 𝑑𝑥
0 Delhi, India, 1984.
𝛼 [7] G. Rahman, S. Mubeen, A. Rehman, and M. Naz, “k-Gamma
𝑘 (77)
= (1 − 𝑒−𝛽𝑥 /𝑘
) ,
𝑥 > 0, 𝛼 > 0, 𝛽 > 0.
The survival function 𝑆(𝑥, 𝛼, 𝛽, 𝑘) is given by
𝑆 (𝑥, 𝛼, 𝛽, 𝑘) = 1 − 𝐹 (𝑥, 𝛼, 𝛽, 𝑘)
𝑘 𝛼
= 1 − (1 − 𝑒−𝛽𝑥 /𝑘
) , 𝑥 > 0.
The hazard rate ℎ(𝑥, 𝛼, 𝛽, 𝑘) is given by
𝐹 (𝑥, 𝛼, 𝛽, 𝑘)
𝑆 (𝑥, 𝛼, 𝛽, 𝑘) =
𝑆 (𝑥, 𝛼, 𝛽, 𝑘)
𝑘 𝛼−1
(1 − 𝑒−𝛽𝑥 /𝑘
) 𝑥𝑘−1 𝑒−𝛽𝑥
= 𝑘 /𝑘 𝛼 , 𝑥 > 0.
1 − (1 − 𝑒−𝛽𝑥 ) [13] S. Mubeen, M. Naz, A. Rehman, and G. Rahman, “Solutions
Similarly we can also obtain the distribution function, relia-
bility function, and hazard rate of other theorems too.
9. Conclusions
In this paper the authors conclude the following.
(i) Weibull distribution is a special case of 𝑘-Generalized
Exponential Distribution. Further if 𝛿 tends to 0, then our
lastly proved more general result leads to Weibull distribu-
tion.
(ii) Also if 𝛿 tends to 1, then our newly introduced 3-
parameter extension of Generalized Exponential Distribu-
tion (GED) reduced to classical Exponential Distribution.
(iii) The moment generating functions (MGFs) obtained
in this paper generalize the classical moment generating
functions (MGFs) of the given distributions.
Conflicts of Interest
The authors declare that they have no conflicts of interest.
References
[1] N. Balakrishnan and A. P. Basu, Exponential Distribution
Theory, Methods and Applications, John Wiley, New York, NY,
USA, 1995.
[2] T. Kondo, “A theory of the sampling distribution of standard
deviation,” Journal of Biometrika, vol. 22, 1/2, pp. 34–64, 1930.
[3] J. F. Steffensen, “Some recent research in the theory of statistics
and actuarial science,” Institute of Actuaries, 1930.
[4] G. Teissier, Dysharmonies et Discontinuités dans la Croissance,
Act. Sci. et Industr., No. 95. (Exposés de Biometrie, 1), Hermann,
Paris, 1934.
[5] W. Weibull, A statistical theory of the strength of ma-terials, vol.
151, Ingeniors Vetenskaps Akademiens, Stockholm, Sweden,
1939.
[6] S. C. Malik, Mathematical analysis, Wiley Eastern Limited, New
and k-Beta distributions and moment generating functions,”
Journal of Probability and Statistics, vol. 2014, Article ID 982013,
6 pages, 2014.
[8] M. G. Kendall and A. Stuart, The Advanced Theory of Statistics,
vol. 2, Charles Griffin and Company, London, UK, 1961.
[9] S. Ross, A First Course in Probability, University of Southern
California, 8th edition.
(78) [10] C. G. Kokologiannaki, “Properties and inequalities of general-
ized 𝑘-gamma, beta and zeta functions,” International Journal of
Contemporary Mathematical Sciences, vol. 5, no. 13–16, pp. 653–
660, 2010.
[11] S. Mubeen, M. Naz, and G. Rahman, “A note on 𝑘-
hypergemetric differential equations,” Journal of Inequalities
and Special Functions, vol. 4, no. 3, pp. 38–43, 2013.
[12] S. Mubeen, G. Rahman, A. Rehman, and M. Naz, “Contiguous
𝑘 (79) function relations for k-hypergeometric functions,” ISRN Math-
/𝑘
ematical Analysis, vol. 2014, Article ID 410801, 6 pages, 2014.
of k-hypergeometric differential equations,” Journal of Applied
Mathematics, vol. 2014, Article ID 128787, 13 pages, 2014.
[14] F. Merovci, “Power product inequalities for the Γk function,”
International Journal of Mathematical Analysis, vol. 4, no. 21, pp.
1007–1012, 2010.
[15] S. Mubeen, A. Rehman, and F. Shaheen, “Properties of 𝑘-
gamma, 𝑘-beta and 𝑘-functions,” Bothalia Journal, vol. 4, pp.
371–379, 2014.
[16] A. Rehman, S. Mubeen, N. Sadiq, and F. Shaheen, “Some
inequalities involving k-beta functions with applications,” Jour-
nal of Inequalities and Applications, 2014:224, 16 pages, 2014.
[17] G. S. Mudholkar and D. K. Srivastava, “Exponentiated Weibull
family for analyzing bathtub failure-rate data,” IEEE Transac-
tions on Reliability, vol. 42, no. 2, pp. 299–302, 1993.
[18] R. D. Gupta and D. Kundu, “Exponentiated exponential family:
an alternative to gamma and Weibull distributions,” Dept. of
Math., Stat. & Comp. Sci, University of New Brusnwick, Saint-
John, Canada, 1997.
[19] R. C. Gupta, P. L. Gupta, and R. D. Gupta, “Modeling fail-
ure time data by Lehman alternatives,” Communications in
Statistics—Theory and Methods, vol. 27, no. 4, pp. 887–904, 1998.
[20] M. P. Yeates, B. J. Tolkamp, D. J. Allcroft, and I. Kyriazakis, “The
use of mixed distribution models to determine bout criteria for
analysis of animal behaviour,” Journal of Theoretical Biology, vol.
213, no. 3, pp. 413–425, 2001.
[21] C. Escalante-Sandoval, “Design rainfall estimation using expo-
nentiated and mixed exponentiated distributions in theCoast of
Chiapas,” Ingenieria Hidraulica en Mexico, vol. 22, pp. 103–113,
2007.
[22] M. T. Madi and M. Z. Raqab, “Bayesian prediction of rainfall
records using the generalized exponential distribution,” Envi-
ronmetrics, vol. 18, no. 5, pp. 541–549, 2007.
[23] R. Subburaj, G. Gopal, and P. K. Kapur, “A software reliability
growth model for vital quality metrics,” The South African
Journal of Industrial Engineering, vol. 18, pp. 93–108, 2007.
Journal of Probability and Statistics 9

[24] F. Biondi, T. J. Kozubowski, A. K. Panorska, and L. Saito, “A new

stochastic model of episode peak and duration for eco-hydro-
climatic applications,” Ecological Modelling, vol. 211, no. 3-4, pp.
383–395, 2008.
[25] J. E. Cota-Felix, F. Rivas-Davalos, and S. Maximov, “An alter-
native method for estimating mean life of power system
equipment with limited end-oflife failure data,” in Proceedings
of the 2009 IEEE Bucharest PowerTech: Innovative Ideas Toward
the Electrical Grid of the Future, July 2009.
[26] D. Kundu and B. Pradhan, “Bayesian inference and life test-
ing plans for generalized exponential distribution,” Science in
China. Series A. Mathematics, vol. 52, no. 6, pp. 1373–1388, 2009.
[27] S. Nadarajah and S. Kotz, “The exponentiated type distribu-
tions,” Acta Applicandae Mathematicae, vol. 92, no. 2, pp. 97–111,
2006.
Advances in Advances in Journal of Journal of
Operations Research
Hindawi Publishing Corporation
Decision Sciences
Hindawi Publishing Corporation
Applied Mathematics
Hindawi Publishing Corporation
Algebra
Hindawi Publishing Corporation
Probability and Statistics
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014

The Scientific International Journal of

World Journal
Hindawi Publishing Corporation
Differential Equations
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014

Submit your manuscripts at

https://www.hindawi.com

International Journal of Advances in

Combinatorics
Hindawi Publishing Corporation
Mathematical Physics
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014

Journal of Journal of Mathematical Problems Abstract and Discrete Dynamics in

Complex Analysis
Hindawi Publishing Corporation
Mathematics
Hindawi Publishing Corporation
in Engineering
Hindawi Publishing Corporation
Applied Analysis
Hindawi Publishing Corporation
Nature and Society
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014

International
Journal of Journal of
Mathematics and
Mathematical
#HRBQDSDĮ,@SGDL@SHBR
Sciences

Journal of International Journal of Journal of

Hindawi Publishing Corporation Hindawi Publishing Corporation Volume 2014

http://www.hindawi.com
Function Spaces
Hindawi Publishing Corporation
Volume 201 http://www.hindawi.com http://www.hindawi.com
Stochastic Analysis
Hindawi Publishing Corporation
Optimization
Hindawi Publishing Corporation
Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014