STATISTICA, anno LXXXIII, n.

4, 2023

DEVELOPMENT AND ESTIMATION OF WEIGHTED

XGAMMA EXPONENTIAL DISTRIBUTION WITH
APPLICATIONS TO LIFETIME DATA

Abhimanyu Singh Yadav

Department of Statistics, Banaras Hindu University, Varanasi, India
Shivanshi Shukla
Department of Statistics, Central University of Rajasthan, Ajmer, India
Neha Jaiswal
Department of Statistics, Banaras Hindu University, Varanasi, India
Sanjay Kumar Singh
Department of Statistics, Banaras Hindu University, Varanasi, India
Debayan Koley 1
Department of Statistics, Banaras Hindu University, Varanasi, India

SUMMARY

In this article, we introduce a weighted version of the Xgamma exponential distribution, extend-
ing its utility in modeling lifetime data. We derive several important distributional properties of
the proposed model, including moments, residual life functions, generating functions, stochastic
ordering, aging intensity, and entropy. These properties provide deeper insights into the behav-
ior and structure of the proposed distribution. To estimate the model parameters, we discuss the
maximum likelihood estimation approach, focusing on complete sample data. To demonstrate
the practical applicability of the proposed distribution, we analyze two real-world lifetime data
sets. The performance of the weighted Xgamma exponential distribution is compared with several
well-established one- and two-parameter lifetime distributions, along with their weighted versions.
Additionally, comparisons are made with length-biased and area-biased lifetime distributions to
further assess the robustness of the proposed model. The results of these comparisons indicate
that the proposed weighted distribution offers a superior fit, particularly for data sets exhibiting
an increasing failure rate. The model’s ability to outperform competing distributions highlights
its potential as an effective alternative for analyzing lifetime data in reliability and survival studies.

Keywords: Xgamma exponential distribution; Weighted Xgamma; Characterizations; Point and

interval estimation.

1
Corresponding Author. E-mail: devstat96@gmail.com
248 A.S. Yadav, S. Shukla, N. Jaiswal, S.K. Singh and D. Koley

1. INTRODUCTION

Introducing new probability distributions is a significant area of research in probability

theory and statistical modeling, as it allows for creating models that can better describe
and capture the complexities of real-world phenomena. Probability distributions are es-
sential in characterizing the behavior of random variables, making them indispensable
in fields like engineering, economics, biology, and social sciences. Developing new dis-
tributions often arises from the need to extend classical models or create new ones that
address the limitations of existing distributions. Classical probability distributions such
as the exponential, gamma, Weibull distributions, etc., have been widely used due to
their simplicity and tractable mathematical properties. However, they may not always
provide the best fit for real-world data. For example, in many applications, data may
exhibit skewness, kurtosis, heavy tails, or multimodal behavior that traditional distri-
butions fail to capture. In such cases, introducing new probability distributions becomes
essential to provide more flexibility and accuracy in modeling. New lifetime distribu-
tions can be derived in several ways. One common method is through generalization or
modification of existing distributions. For instance, a researcher may extend any base-
line distribution to account for skewness or use transformations to handle heavy-tailed
data. A common approach to introducing new lifetime distributions includes general-
izing existing distributions, where classical models like the Weibull or exponential are
extended by adding parameters or transformations to increase flexibility. Another ap-
proach is compounding or mixing distributions, where two or more distributions are
combined to capture more complex data patterns. Parameterized extensions introduce
additional shape, scale, or skewness parameters to enhance model flexibility. Generat-
ing families of distributions creates broad classes of models, such as exponentiated or
Kumaraswamy families.

The motivation of our present study is to introduce a new-sprung weighted distribu-

tion and further explore various distributional characteristics. The need for the study of
weighted distribution arises because when a sample is gathered by the experimenter fol-
lowing some standard probability model, the obtained sample may not represent the ac-
tual stochastic model considered due to accidental loss, destruction, or non-observance
of some units until and unless each unit is selected with an equal chance of inclusion in
the sample. Therefore, to quench the thirst for fitting our data in a better way, we resort
to weighted distributions. Weighted distributions try to obtain the specified probabili-
ties of the events/sample being observed and recorded by adjusting the probabilities of
the actual occurrence of the events/sample. Mathematically, Consider a finite or infinite
population of units carrying values of a non-negative random variable X having a distri-
bution with probability density function (PDF) g (x) where x ≥ 0. Usually, we consider
that the probability of selection of each unit is the same, regardless of the value of x it
carries so that the PDF for the observation x is g (x); but if the respective probability of
observing x is 0 < w(x) < 1, then the corresponding PDF of the recorded observation
Development and Estimation of Weighted Xgamma Exponential Distribution 249

is
g (x)w(x)
gw (x) = R . (1)
x
g (x)w(x)d x

The above defined PDF emulates g (x) only when w(x) is constant.
The nascent idea of weighted distribution is attributed to Fisher (1934) who threw
light on the influence of methods of ascertainment while approximating the frequencies.
The idea was further strengthened by Rao (1965). He illustrated the practicality of the
weighted distribution when w(x) = x or w(x) = x a . When the weight merely corre-
sponds to the size of the respective sampling unit then the weighted scheme is said to
be size-biased and is discussed by Patil and Ord (1976). Size-biased distribution is a cele-
brated version of weighted distribution and is further characterized as length-biased and
area-biased distribution. Patil and Rao (1978) further discussed the concept of weighted
distributions in the realm of human populations and wildlife management. Warren and
Olsen (1964) coined the term ‘line-intersect sampling’ in the context of forestry litera-
ture. Grosenbaugh (1958) studied the concept and applicability of area sampling in the
forestry regime. Lele and Keim (2006) used the theory of weighted distributions to es-
timate the resource selection probability function, which is defined by the probability
of an individual or organism selecting a resource available to their surrounding envi-
ronment. Larose and Dey (1998) modeled publication bias using weighted distributions
under the Bayesian framework. Gupta and Keating (1986) established the relationships
among various reliability measures of a lifetime distribution under length-bias sampling.
Jain et al. (1989) generalized those results for a weighted lifetime distribution. Sunoj and
Maya (2006) discussed some properties of weighted distributions in the context of re-
pairable systems. It is to be noted that while using the weighted distribution for better
fitting of recorded observations, appropriate choice of weight is immensely important.
The primary objective of this article is to introduce and thoroughly examine the
various distributional properties of a newly proposed weighted distribution, along with
its specific cases, namely the length-biased and area-biased distributions. In addition to
studying these properties, a key aim is to develop and propose maximum likelihood
(ML) estimation procedures for accurately estimating the unknown parameters and de-
termining the survival/reliability characteristics of the proposed distribution. The re-
search further aims to evaluate the performance and efficiency of these proposed es-
timators for different sample sizes and varying model parameters using Monte Carlo
simulation techniques. Through these simulations, the robustness and accuracy of the
estimation methods will be assessed, offering insight into how well the estimators per-
form under different conditions. Furthermore, the article intends to demonstrate the
practical relevance and applicability of the proposed distribution by evaluating its fit
with real-world data sets.
To attain our first objective, using Xgamma exponential distribution (XGED) as the
baseline model, we study the size-biased (weighted) distribution with weight w(x) = x υ .
For υ = 1 and υ = 2 we get length-biased and area-biased weighted distributions respec-
tively. In survival/reliability analysis, many models are available to describe the lifetime
250 A.S. Yadav, S. Shukla, N. Jaiswal, S.K. Singh and D. Koley

data and to explore the inherent characteristics of the complete/ censored data. These
lifetime models are classified based on their hazard rate. The one-parameter exponen-
tial distribution is a very celebrated survival/reliability model due to its constant hazard
rate and lack of memory property. Since its application is restricted only to the constant
hazard rate, therefore several generalizations based on the exponential model have been
developed and characterized through non-constant hazard rates. XGED model is one of
these which was introduced by Yadav et al. (2022) as an alternative to the one-parameter
exponential distribution. The beauty of this distribution has been fully justified based
on its monotone and bathtub-shaped hazard rate.
The cumulative distribution function (CDF) and corresponding PDF of XGED is
given by
1 −ζ x 1
• ˜
2
FXGED (x; ζ ) = 1 − e 2 + ζ x + (ζ x) , x > 0, ζ > 0 (2)
2 2
and
1 1
• ˜
fXGED (x; ζ ) = ζ e−ζ x 1 + (ζ x)2 , x > 0, ζ > 0, (3)
2 2
respectively. The hazard function for the above distribution is given as
1
” —
ζ 1 + 2 (ζ x)2
hXGED (x; ζ ) = 1
. (4)
2 + ζ x + 2 (ζ x)2

This current study is motivated due to the variety of applications of the weighted dis-
tribution in lifetime data analysis. To the best of our knowledge, no attempt has been
made in this direction, enhancing the uniqueness of the proposed study.
The article is structured as follows. Section 1 provides an overview of the relevant
literature and outlines the motivation for the study. In Section 2, the newly proposed
model is introduced, along with an in-depth examination of its various statistical prop-
erties. Section 3 details the point and interval estimation procedures for the model pa-
rameters and the survival/reliability characteristics, while Section 4 evaluates the per-
formance of the classical estimators for distribution parameters and survival/reliability
characteristics through a Monte Carlo simulation. In Section 5, the practical application
of the proposed model is demonstrated using a real-world reliability and survival data
set. Finally, Section 6 concludes the paper with a summary of key findings and remarks.

2. THE PROPOSED DISTRIBUTION AND CHARACTERIZATION

In this Section, we introduce the weighted Xgamma exponential distribution (WXGED)

by considering the weight function as w(x) = x υ . The PDF and CDF with scale param-
eter ζ > 0 and shape parameter υ > 0 can be formally defined as
(ζ x)2
2ζ υ+1 x υ e −(ζ x) [1 + 2
]
gw (x; υ, ζ ) = , x >0 (5)
υ!(υ2 + 3υ + 4)
Development and Estimation of Weighted Xgamma Exponential Distribution 251

PDF plot for PDF plot for

PDF plot for WXGED
length−biased WXGED area−biased WXGED
0.5

0.5

0.5
υ = 0.5 , ζ = 0.25 υ = 1 , ζ = 0.25 υ = 2 , ζ = 0.25
υ = 0.75 , ζ = 0.5 υ = 1 , ζ = 0.5 υ = 2 , ζ = 0.5
υ = 1.25 , ζ = 0.75 υ = 1 , ζ = 0.75 υ = 2 , ζ = 0.75
υ = 1.5 , ζ = 1 υ=1,ζ=1 υ=2,ζ=1
υ = 2.5 , ζ = 1.25 υ = 1 , ζ = 1.25 υ = 2 , ζ = 1.25
0.4

0.4

0.4
υ=3,ζ=2 υ=1,ζ=2 υ=2,ζ=2
0.3

0.3

0.3
density

density

density
0.2

0.2

0.2
0.1

0.1

0.1
0.0

0.0

0.0
0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
Observations Observations Observations
(a) (b) (c)

Figure 1 – PDF plot of WXGED for different υ and ζ values: (a) WXGED(υ, ζ ) (b) WXGED(υ =
1, ζ ) and (c) WXGED(υ = 2, ζ ).

and
[2IL (υ + 1, ζ x) + IL (υ + 3, ζ x)]
Gw (x; υ, ζ ) = , x > 0, (6)
υ!(υ2 + 3υ + 4)
respectively, where IL denotes the lower incomplete gamma integral2 . The model pro-
posed above produces length-biased and area-biased WXGED for υ = 1 and υ = 2 re-
spectively. Putting υ = 1 and υ = 2 in Equations (5) and (6), we get PDF and CDF of
length-biased and area-biased WXGED respectively. The shape of the PDF for differ-
ent combinations of parameters (υ, ζ ) is displayed in Figure 1. The PDF is also plotted
in particular for υ = 1 (length-biased), and υ = 2 (area-biased). From those figures, it is
observed that the PDF is positively skewed. The two most important survival character-
istics are survival function and hazard rate, which, in the case of WXGED, are defined
in the following equations:

[2IU (υ + 1, ζ t ) + IU (υ + 3, ζ t )]
Sw (t ; υ, ζ ) = , t >0 (7)
υ!(υ2 + 3υ + 4)

and
(ζ t )2
” —
2ζ υ+1 t υ e −ζ t 1 + 2
hw (t ; υ, ζ ) = , t > 0, (8)
[2IU (υ + 1, ζ t ) + IU (υ + 3, ζ t )]

Rx
The lower incomplete gamma integral IL (a, x) is defined as 0 t a−1 exp(−t )d t and the upper
2
R ∞ a−1
incomplete gamma integral IU (a, x) is defined as x t exp(−t )d t , such that IL (a, x)+IU (a, x) =
Γ (a).
252 A.S. Yadav, S. Shukla, N. Jaiswal, S.K. Singh and D. Koley

respectively, where IU denotes upper incomplete gamma integral. Note that, from a
theoretical point of view, the survival function is analogous to the reliability function
and has similar interpretations.
Now, the mode of the distribution, which is nothing but the observation(s) with the
highest frequency, can be obtained as a solution of gw′ (x; υ, ζ ) = 0, where
 
υ ζ 2
 
d x
gw′ (x; υ, ζ ) = gw (x; υ, ζ ) log gw (x; υ, ζ ) = gw (x; υ, ζ )  − ζ + .
d (x) x 1+
(ζ x)2
2
(9)
Since gw (x; υ, ζ ) ̸= 0, therefore
 
υ
 −ζ + ζ 2
x  = 0.
(ζ x)2
(10)
x 1+ 2

After simplification we get,

ζ 3 x 3 − ζ 2 (υ + 2) x 2 + 2ζ x − 2υ = 0.
 
(11)

The above equation can be easily solved by some numerical method which will finally
provide us with the expression for mode.

2.1. Shape of hazard rate

The shape of the hazard rate of the proposed distribution is abstruse at first sight as it
involves an incomplete gamma function in its expression. Therefore, we resort to the
lemma given by Glaser (1980) which states the implication of monotonicity between
g ′ (t )
η(t ) = − g (t ) and h(t ). Using the result, the shape of the hazard rate can be deduced as
follows.

PROPOSITION 1. The hazard function of the WXGED(υ, ζ ) is an increasing function

of t for given ζ > 0 and υ > 0.

PROOF. for T ∼ WXGED(υ, ζ ), η(t ) is given by

υ 2ζ 2 t
η(t ) = ζ − − . (12)
t 2 + (ζ t )2

Consequently,
4(υ − 1)ζ 2 t 2 + ζ 4 t 4 (υ + 2) + 4υ
η′ (t ) = . (13)
t 2 (2 + ζ 2 t 2 )2
Since ζ , υ, t all are greater than zero, therefore η′ (t ) > 0, ∀ t > 0. 2
Development and Estimation of Weighted Xgamma Exponential Distribution 253

Hazard plot for Hazard plot for

Hazard plot for WXGED
length−biased WXGED area−biased WXGED
3.0

3.0

3.0
υ = 0.5 , ζ = 0.25 υ = 1 , ζ = 0.25 υ = 2 , ζ = 0.25
υ = 0.75 , ζ = 0.5 υ = 1 , ζ = 0.5 υ = 2 , ζ = 0.5
υ = 1.25 , ζ = 0.75 υ = 1 , ζ = 0.75 υ = 2 , ζ = 0.75
υ = 1.5 , ζ = 1 υ=1,ζ=1 υ=2,ζ=1
2.5

2.5

2.5
υ = 2.5 , ζ = 1.25 υ = 1 , ζ = 1.25 υ = 2 , ζ = 1.25
υ=3,ζ=2 υ=1,ζ=2 υ=2,ζ=2
2.0

2.0

2.0
hazard

hazard

hazard
1.5

1.5

1.5
1.0

1.0

1.0
0.5

0.5

0.5
0.0

0.0

0.0
0 10 20 30 40 50 0 10 20 30 40 50 0 10 20 30 40 50
Observations Observations Observations
(a) (b) (c)

Figure 2 – Hazard plot of WXGED for different υ and ζ values: (a) WXGED(υ, ζ ) (b)
WXGED(υ = 1, ζ ) and (c) WXGED(υ = 2, ζ ).

Hence it is deduced that the hazard function of WXGED is an increasing function of t

for given υ and ζ . The shape of the hazard rate for different combinations of parameters
(υ, ζ ) is displayed in Figure 2. The hazard rate function is also plotted in particular
for υ = 1 (length-biased), υ = 2 (area-biased). From those figures, it is observed that
the shape of the failure rate is indeed monotonically non-decreasing. Interestingly, the
increasing hazard rate exhibited by the proposed distribution is highly spotted in the
arena of clinical trials, inventory control, and many other fields in the industrial sector.

2.2. Moments
Moments constitute an important inferential aspect of any model. They form the basis
to comment on measures like the central tendency of the distribution, scatteredness,
asymmetricity, and peakedness. Next, we derive the general expression for the raw mo-
ments of WXGED in the following manner:
Z∞
′ r
µ r (x) = E(X ) = x r gw (x; υ, ζ )d x. (14)
0

After simplification we get,

2(υ + r )! (υ + r + 1)(υ + r + 2)
• ˜
µ′r (x) = 1+ ; r ≥ 1. (15)
ζ r υ!(υ2 + 3υ + 4) 2

Putting υ = 1 and υ = 2, we get the general expression for raw moments in the case of
length-biased and area-biased distribution, respectively.
254 A.S. Yadav, S. Shukla, N. Jaiswal, S.K. Singh and D. Koley

Particularly, mean and variance of WXGED(υ, ζ ) are given by

(υ + 1)(υ2 + 5υ + 8)
µ′1 = (16)
ζ (υ2 + 3υ + 4)

and
υ+1 (υ + 1)(υ2 + 5υ + 8)2
 
2
µ2 = (υ + 2)(υ + 7υ + 14) − , (17)
ζ 2 (υ2 + 3υ + 4) (υ2 + 3υ + 4)

respectively. By substituting υ = 1 and υ = 2 in the above equations, we get a similar

expression for length-biased and area-biased distribution, respectively.
As mean and variance respectively give the measures of central tendency and disper-
sion of a distribution, measures of skewness and kurtosis provide the necessary means
for understanding the shape of a particular distribution in terms of asymmetricity and
peakedness, respectively. To obtain the coefficient of skewness and kurtosis, one needs
to derive the second, third, and fourth central moments. These moments can be calcu-
lated using the expressions for the raw moments given above.
Evaluated expressions of coefficient of skewness and kurtosis for WXGED are
3(υ+1)(υ+2)(υ2 +5υ+8)(υ2 +7υ+14) 2(υ+1)2 (υ2 +5υ+8)3
h i
 1 (υ + 2)(υ + 3)(υ2 + 9υ + 22) − +
υ2 + 3υ + 4

2 (υ2 +3υ+4) (υ2 +3υ+4)2
γ1 = (18)
υ+1 h
(υ+1)(υ2 +5υ+8)2
i3
2
(υ + 2)(υ2 + 7υ + 14) − (υ2 +3υ+4)

and
υ2 + 3υ + 4 [a − b + c − d ]
 
γ2 = — , (19)
υ+1 (υ+1)(υ2 +5υ+8)2 2
”
(υ + 2)(υ2 + 7υ + 14) − (υ2 +3υ+4)

respectively, where

4(υ + 1)(υ + 2)(υ + 3)(υ2 + 9υ + 22)(υ2 + 5υ + 8)

a = (υ+2)(υ+3)(υ+4)(υ2 +11υ+32), b = ,
(υ2 + 3υ + 4)

6(υ + 2)(υ + 1)2 (υ2 + 7υ + 14)(υ2 + 5υ + 8)2 3(υ + 1)3 (υ2 + 5υ + 8)4
c= ,d= .
(υ2 + 3υ + 4)2 (υ2 + 3υ + 4)3

It can be noted that the coefficient of skewness and kurtosis of WXGED are func-
tions of υ. The coefficient of skewness and kurtosis plots for increasing values of υ are
given in Figure 3. From those plots, it can be seen that the WXGED is positively skewed
and leptokurtic. Although, for increasing values of υ, the coefficients of skewness and
kurtosis decrease gradually.
The moment generating function (MGF) M x (t ) is a powerful tool to generate mo-
ments but as it is demarcated by a limited range, we go for evaluating the expression for
Development and Estimation of Weighted Xgamma Exponential Distribution 255
Coefficient of Skewness

1.2

5.0
Coefficient of Kurtosis
1.0

4.5
0.8

4.0
0.6

3.5
0.4

0 5 10 15 20 0 5 10 15 20

υ υ
(a) (b)

Figure 3 – (a) Coefficient of Skewness plot and (b) Coefficient of Kurtosis plot of WXGED for
increasing υ.

the characteristic function Φ x (t ). The beauty of the latter is highlighted by the fact that
it always exists while the former may or may not exist.
The expression of MGF after due simplification is given by

‹υ+1 • ‹2 ˜
1 ζ ζ
 
M x (t ) = 2 + (υ + 1)(υ + 2) . (20)
υ2 + 3υ + 4 ζ − t ζ −t

Characteristic function Φ x (t ) can be obtained by putting i t in M x (t ) in place of t , where

p
i = −1. The respective expressions for length-biased and area-biased distribution can
be obtained by putting υ = 1 and υ = 2. Further, if the life of the unit x under obser-
vation exceeds t ′ ; one may seek the expressions for moments and moment-generating
functions in that situation. Evaluated expressions are termed conditional moments and
conditional MGFs, respectively. Mathematically the conditional moments and the con-
ditional MGF are obtained as
R∞ R∞
r ′ t′
x r gw (x; υ, ζ )d x tX ′ t′
e t x gw (x; υ, ζ )d x
E[X |X > t ] = and E[e |X > t ] = ,
P (X > t ′ ) P (X > t ′ )
(21)
respectively. After substituting the values from Equation (5) and Equation (7) in Equa-
256 A.S. Yadav, S. Shukla, N. Jaiswal, S.K. Singh and D. Koley

tion (21), the expressions for conditional moments and MGF are given as

2IU (υ + r + 1, ζ t ′ ) + IU (υ + r + 3, ζ t ′ )
E[X r |X > t ′ ] = (22)
ζ r [2IU (υ + 1, ζ t ′ ) + IU (υ + 3, ζ t ′ )]

and
2ζ υ+1 ζ υ+3
I (υ + 1, (ζ
(ζ −t )υ+1 U
− t )t ′ ) + (ζ −t )υ+3 IU (υ + 3, (ζ − t )t ′ )
E[e t X |X > t ′ ] = , (23)
2IU (υ + 1, ζ t ′ ) + IU (υ + 3, ζ t ′ )

respectively. For the corresponding expressions in the case of length-biased and area-
biased distribution put υ = 1 and υ = 2 in Equations (22) and (23), respectively.
REMARK 2. With t ′ → 0; E(X r |X > t ′ ) → E(X r ) and E(e t X |X > t ′ ) → E(e t X ),
respectively.

2.3. Mean residual lifetime

The beauty of the mean residual lifetime (MRL) function lies in the fact that it summa-
rizes the entire residual life distribution. It holds immense importance in the fields of
demography, actuarial sciences, and different areas of the industrial sector. MRL func-
tion e(x) of a non-negative random variable X with survivor function S(x) is given as

e(x) = E(X − t |X > t ) (24)

R∞
t
y gw (y; υ, ζ )d y
= − t.
Sw (t ; υ, ζ )

After the required calculations we get

2IU (υ + 2, ζ t ) + IU (υ + 4, ζ t )
e(x) = − t. (25)
2ζ IU (υ + 1, ζ t ) + ζ IU (υ + 3, ζ t )

To obtain the due expressions for length-biased and area-biased distribution put υ = 1
and υ = 2 in the above equations, respectively.

2.4. Residual lifetime distribution

The residual lifetime R(t ) is the additional life of a unit with the assertion that it has
survived up to time t , i.e. R(t ) = X − t |X > t , t > 0 whereas reverse residual lifetime
R̄(t ) is the time to observe the failure given that the unit has life less than or equal to t
i.e; R̄(t ) = t − X |X ≤ t , t > 0. The survival function for R(t ) is given by

Ḡ(x + t )
SR(t ) = , (26)
Ḡ(t )
Development and Estimation of Weighted Xgamma Exponential Distribution 257

where Ḡ(·) is the survival function. Substituting values from Equation (7) in Equation
(26) we get
2I (υ + 1, ζ (x + t )) + IU (υ + 3, ζ (t + x))
SR(t ) = U . (27)
2IU (υ + 1, ζ t ) + IU (υ + 3, ζ t )

Now, the expressions for R̄(t ), t > 0 are obtained in similar way. The corresponding
PDF and the hazard function for R(t ) and R̄(t ) can be derived by using the relation
′ f (t )
f (t ) = −S (t ) and h(t ) = S(t ) , respectively. By putting υ = 1 and υ = 2 in the above
equation respective expressions for length-biased and area-biased distribution are ob-
tained.

2.5. Stochastic ordering

Stochastic ordering has proved to be highly useful in many diverse areas of probability
and statistics, especially in the financial sector. In a financial setting, stochastic orders
help to decide on the maximum return subject to a given utility function. The detailed
description of stochastic ordering can be found in Shaked and Shanthikumar (1994). The
following theorem shows the stochastic ordering of two random variables X and Y in
the context of likelihood ratio order.

THEOREM 3. Let X and Y be two independent random variables that follow WXGED
with shape parameters υ1 and υ2 and scale parameters ζ1 and ζ2 respectively. If υ1 > υ2 and
ζ2 > ζ1 , then X is stochastically greater than Y in likelihood ratio order, i.e. (Y ≤ l r X ) for
all x.

PROOF. For given X ∼ WXGED(υ1 , ζ1 ) and Y ∼ WXGED(υ2 ,ζ2 ), we have

gw (x; υ1 , ζ1 )
ψ = ,
gw (x; υ2 , ζ2 )
 υ1 −υ2 ζ 3 x2 ζ 3 x2
ζ1 υ2 !(υ22 + 3υ2 + 4)
 ‹
υ1 −υ2 −x ζ1 + 2 −ζ2 − 2
1 2
ψ = x e (28)
ζ2 υ1 !(υ21 + 3υ1 + 4)

and
υ 1 − υ2  3x 2 3
 
d
ψ=ψ − (ζ1 − ζ2 ) − ζ1 − ζ23 .


(29)
dx x 2
The above equation increases in x for all υ1 > υ2 and ζ2 > ζ1 . 2

COROLLARY 4. Let X ∼ WXGED(υ1 , ζ1 ) and Y ∼ WXGED(υ2 ,ζ2 ). If υ1 > υ2 and

ζ2 > ζ1 then X is stochastically greater than Y in likelihood ratio order and the result also
holds in hazard rate order and mean residual life order.
258 A.S. Yadav, S. Shukla, N. Jaiswal, S.K. Singh and D. Koley

AI plot for AI plot for

AI plot for WXGED
length−biased WXGED area−biased WXGED

2.0

3.5
4

υ = 0.5 , ζ = 0.25
υ = 0.75 , ζ = 0.5
υ = 1.25 , ζ = 0.75

3.0
υ = 1.5 , ζ = 1
υ = 2.5 , ζ = 1.25

1.8
υ=3,ζ=2
3

2.5
1.6

2.0
AI

AI

AI
2

1.5
1.4
υ = 2 , ζ = 0.25
υ = 2 , ζ = 0.5

1.0
υ = 1 , ζ = 0.25 υ = 2 , ζ = 0.75
1

υ = 1 , ζ = 0.5 υ=2,ζ=1
1.2

υ = 1 , ζ = 0.75 υ = 2 , ζ = 1.25
υ=1,ζ=1 υ=2,ζ=2

0.5
υ = 1 , ζ = 1.25
υ=1,ζ=2
1.0

0.0
0

0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
Observations Observations Observations
(a) (b) (c)

Figure 4 – AI plot of WXGED for different υ and ζ values: (a) WXGED(υ, ζ ) (b) WXGED(υ =
1, ζ )and (c) WXGED(υ = 2, ζ ).

COROLLARY 5. It can also be noted that, as shown in Jain et al. (1989), if we choose our
weight function w(x) as a monotonically increasing function of x, then the weighted version
of the distribution dominates the original distribution in likelihood ratio order. Since our
weight function x υ is indeed a monotonically increasing function of x, random variables
from WXGED are stochastically greater than random variables from XGED in likelihood
ratio order.

2.6. Aging intensity

Aging is very crucial and an inherent property of a system (may be living or non-living)
and is obtained using a hazard rate. Generally, it is defined as a ratio of the failure rate
to the average hazard rate. For a given failure rate, there exists a unique aging intensity
(AI) function but not the other way around. Further, it can be noted that the monotonic
behavior of the failure rate function does not ensure the same monotonic behavior of
the AI function. For more details about AI, see Nanda et al. (2007). AI is expressed in
the following formula:

−t g (t )
L x (t ) = ; t > 0. (30)
S(t ) log S(t )
Development and Estimation of Weighted Xgamma Exponential Distribution 259

In the case of the proposed model, the expression for AI is given by

(ζ t )2
€ Š
2t υ+1 ζ υ+1 e −ζ t 1 + 2
L x (t ) = , (31)
S (t ; υ, ζ )

where
S (t ; υ, ζ ) = [2IU (υ + 1, ζ t ) + IU (υ + 3, ζ t )][log υ! + log(υ2 + 3υ + 4)
− log (2IU (υ + 1, ζ t ) + IU (υ + 3, ζ t ))].

The shape of the AI function for different combinations of parameters (υ, ζ ) is displayed
in Figure 4. The AI function is also plotted in particular for υ = 1 (length-biased), υ = 2
(area-biased). From those figures, it is observed that the shape of the AI function in-
creases for some observations, but at the beginning and the end of the curve, it decreases.

2.7. Measure of uncertainty

Renyi entropy is immensely useful in the fields of statistical inference, econometrics,
and pattern recognition in computer science for measuring the uncertainty associated
with the phenomena. For WXGED, the expression for the Renyi entropy function is
given by
(ζ x)2
!ν
2ζ υ+1 x υ e −(ζ x) [1 + 2 ]
Z∞ Z∞
1 ν 1
RE(ν) = log gw (x; υ, ζ ) = log d x.
1−ν 0 1−ν 0 υ!(υ2 + 3υ + 4)

After simplification, the final expression of Renyi entropy is obtained as

ν ν    2 i
2ζ υ+1 (υν + 2i)!
– ™
1 X ν ζ
RE(ν) = log . (32)
1−ν υ!(υ2 + 3υ + 4) i =0 i 2 (ζ ν)υν+2i +1

2.8. Bonferroni and Lorenz curve

Bonferroni curve and Lorenz curve are two popular indices used in the financial arena,
particularly in the context of income inequality. These two measures have some appli-
cations in reliability and life-testing experiments as well. The Bonferroni curve is given
by the following expression:

G1 (x)
BG ( p) = ,
G(x)

where
1 Rx
G1 (x) = µ 0
t g (t )d t , µ = E(X ).
260 A.S. Yadav, S. Shukla, N. Jaiswal, S.K. Singh and D. Koley

Putting p = G(x) and q = G −1 ( p) we get

Zq
1
BG ( p) = t g (t )d t , p ∈ (0, 1].
pµ 0

The Lorenz curve is simply given by

LG ( p) = pBG ( p).

The evaluated expressions for these two curves in the case of WXGED are as follows:

2 1
• ˜
BG ( p) = IL (υ + 2, ζ q) + IL (υ + 4, ζ q) , (33)
pζ µυ!(υ2 + 3υ + 4) 2
2 1
• ˜
LG ( p) = IL (υ + 2, ζ q) + IL (υ + 4, ζ q) .
µζ υ!(υ2 + 3υ + 4) 2

3. MAXIMUM LIKELIHOOD ESTIMATION

Let X1 , X2 , . . . , Xn be a random sample of size n coming from WXGED with PDF and
CDF defined in Equations (5) and (6), respectively. The log-likelihood (LL) function for
the proposed distribution is given by
n
X
log L(ζ , υ|x ) = n log(2) − n log(υ!) − n log(υ2 + 3υ + 4) + n(υ + 1) log ζ − ζ xi
e i=1
(34)
X n Xn ζ 2 xi2
+υ log xi + log(1 + ).
i =1 i=1
2

Differentiating log L(ζ , υ|x ) with respect to υ, ζ and equating differentials to zero the
following score equations eare obtained
υ n
n(2υ + 3)
–‚ Œ™
X 1 X
−n −γ + − + n log(ζ ) + log xi = 0 (35)
k=1
k (υ2 + 3υ + 4) i=i

and

n(υ + 1) X n X n ζ xi2
− xi + = 0, (36)
ζ i=1 i =1 (1 +
ζ 2 xi2
)
2

where γ = Euler-Mascheroni constant.

Solving Equations (35) and (36) for υ and ζ , respectively, we get the ML estimators
of the parameters. The solutions for the above score equations can not be obtained ana-
lytically because of their complicated expressions. Therefore, numerical procedures like
Development and Estimation of Weighted Xgamma Exponential Distribution 261

the Newton-Raphson algorithm or fixed-point iterations are used to find the solutions
to these equations.
Further, one may be interested in estimating the survival and hazard functions de-
fined in Equations (7) and (8), as it conveys meaningful information about the
survival/reliability properties of the lifetime distribution. The invariance property of
the ML estimator allows us to directly obtain the ML estimators of the survival and haz-
ard function by simply plugging in the ML estimators of the distribution parameters.
Hence, if υ̂ and ζˆ denote the ML estimators of υ and ζ , respectively, then the ML esti-
mators of the survival and hazard function of WXGED is defined as Ŝ(t ) = Sw (t ; υ̂, ζˆ)
and ĥ(t ) = h (t ; υ̂, ζˆ), respectively.
w
The 100(1 − α)% asymptotic confidence intervals (ACI) for the distribution param-
eters υ and ζ can be constructed using the asymptotic normality property of the ML
estimators, which are well-discussed in the literature. The ACIs for the survival and haz-
ard functions are also available through the delta method, which was beyond the scope
of this study.

4. SIMULATION STUDY

In this Section, a Monte Carlo simulation study has been conducted to evaluate the
performance of the ML estimators for the parameters and survival/reliability character-
istics of the proposed distribution. The study is carried out based on different combina-
tions of distributional parameters, specifically (1, 1.75), (2, 2.5) and (2.75, 3). To provide
a comprehensive comparison among the estimators, the first two parameter combina-
tions where ν = 1, 2 correspond to the length-biased and area-biased distributions, re-
spectively, and the last parameter combination indicates a size-biased (weighted) distri-
bution, allowing us to explore the effect of different biasing schemes on the estimators’
performance. Further, the ML estimators of the survival and the hazard function are
evaluated at an arbitrary point t = 4. To assess the performance of the estimators un-
der different conditions, five sample sizes are chosen: n = 10, 20, 30, 50, and 100. The
accuracy of the estimators is measured in terms of bias and mean squared error (MSE).
Once the sequence of samples is generated, the average estimate (AE), bias, and MSE are
calculated for each parameter combination, sample size, and biasing scheme. The AE,
bias, and MSEs for the considered parameters and survival/reliability characteristics are
obtained by calculating using the following formula based on N = 10000 replications:
N N N
1 X 1 X 1 X
AE = Θ̂i , Bias = (Θ̂i − Θ) and MSE = (Θ̂ − Θ)2 ,
N i =1 N i=1 N i=1 i

where Θ = (υ, ζ , Sw (t ; υ, ζ ), hw (t ; υ, ζ )).

In addition to point estimation, ACIs for the parameters υ and ζ are constructed
for the same variations in distributional parameters and sample sizes considered in the
study. The interval estimators are evaluated in terms of coverage probability (CP) and
262 A.S. Yadav, S. Shukla, N. Jaiswal, S.K. Singh and D. Koley

average width (AW), where

N N
1 X 1 X
CP = I (Li ≤ Θ′ ≤ Ui ) and AW = (U − Li ),
N i=1 N i=1 i

where Θ′ = (υ, ζ ), I (·) is an indicator function, Li is the lower bound and Ui is the upper
bound of the ACIs, respectively.
A lower MSE indicates a better-performing estimator. The results of both the point
estimation and the interval estimation for the different parameter variations and sample
sizes are presented in Table 1. The simulation study revealed several important trends.
First, as the sample size increases, both the MSE and the bias of each estimator decrease
consistently across all parameter combinations. This indicates that larger sample sizes
lead to more accurate and reliable estimates, as the estimators tend to converge to the
true parameter values. In terms of interval estimation, the CP of the ML estimator
approaches the nominal confidence level in all of the cases considered. This means that
the proportion of intervals that contain the true parameter value becomes closer to the
desired confidence level as the sample size grows. Furthermore, the AW of the ACIs
decreases as the sample size increases. This reduction in interval width signifies greater
precision in the interval estimates, as larger sample sizes provide more information about
the parameter, resulting in tighter confidence intervals. Overall, the results show that
as the sample size grows, the estimators become more accurate (with smaller bias and
MSE), and the confidence intervals become both narrower and more reliable.

5. REAL DATA APPLICATIONS

In this Section, two data sets have been taken to show the application of the proposed
model. Data set 1 has been taken from Nichols and Padgett (2006) which represents
the breaking stress of carbon fibers (in GBA). Data set 2 represents the data of lifetime
relating to relief times (in minutes) of 20 patients receiving an analgesic. This data was
originally reported by Gross and Clark (1975) and it was used by Shukla (2019) in a
comparative study on one parameter lifetime distributions. The descriptive statistics of
the two data sets along with the box plot and histogram are given in Table 2, Figure 5,
and Figure 7, respectively. A model-fitting summary of both data sets has been shown
in Tables 3 and 4, respectively.
For Data set 1, we have compared our proposed WXGED model with different dis-
tributions namely weighted exponential distribution (WED), generalized Lindley dis-
tribution (GLD), generalized Exponential distribution (GED), Rayleigh distribution
(RD), XGED, Lindley distribution (LD), exponential distribution (ED), and Xgamma
distribution (XGD). total time on test (TTT) plot for the Data set 1 and empirical CDF
(eCDF) vs Theoretical CDF plots are also given in Figure 6.
Since our proposed model has increasing failure rate properties, only the distribu-
tions with increasing hazard rate properties are considered for comparison, except for
ED, which has a constant failure rate. The motivation for taking ED as an alternative
Development and Estimation of Weighted Xgamma Exponential Distribution 263

TABLE 1
AE, Bias and MSE of different pairs of υ, ζ , corresponding S(t ) and h(t ) and 95% CP and AW for
ACIs for different combinations of n.
S(t )| t =4 h(t )| t =4
n (υ, ζ ) AE Bias MSE CP AW AE Bias MSE AE Bias MSE
1.87107 0.87107 4.39602 0.92270 4.17087
(1, 1.75) 0.06236 -0.00079 0.00284 1.44647 0.31877 0.60573
2.27944 0.52944 1.53162 0.95530 3.10467
10 3.58645 1.58645 14.37851 0.87140 6.64027
(2, 2.5) 0.02752 0.00205 0.00097 2.16374 0.54037 1.72842
3.42748 0.92748 4.79753 0.90490 4.50047
4.85724 2.10724 24.45878 0.84850 8.06914
(2.75, 3) 0.01697 0.00226 0.00048 2.63955 0.70197 2.71868
4.21863 1.21863 7.89245 0.84450 5.27907
1.31947 0.31947 0.78403 0.97330 2.62904
(1, 1.75) 0.06312 -0.00003 0.00155 1.25534 0.12763 0.14592
1.95477 0.20477 0.32489 0.95950 1.87546
20 2.63090 0.63090 2.62429 0.94170 4.60986
(2, 2.5) 0.02626 0.00079 0.00046 1.84269 0.21931 0.35499
2.87259 0.37259 0.92408 0.94050 2.91038
3.59500 0.84500 4.50198 0.90170 5.79289
(2.75, 3) 0.01582 0.00111 0.00022 2.22421 0.28663 0.57853
3.49303 0.49303 1.56442 0.92670 3.55085
1.19989 0.19989 0.40216 0.97330 2.07053
(1, 1.75) 0.06331 0.00016 0.00104 1.20643 0.07873 0.07980
1.87683 0.12683 0.17318 0.94730 1.44828
30 2.38689 0.38689 1.25257 0.95120 3.61597
(2, 2.5) 0.02628 0.00081 0.00032 1.75840 0.13503 0.18619
2.72878 0.22878 0.46220 0.93880 2.26773
3.25413 0.50413 2.13414 0.93090 4.60782
(2.75, 3) 0.01572 0.00101 0.00015 2.10958 0.17200 0.28903
3.29491 0.29491 0.75746 0.93750 2.81262
1.11465 0.11465 0.17786 0.95620 1.55516
(1, 1.75) 0.06302 -0.00013 0.00063 1.17484 0.04713 0.04051
1.82481 0.07481 0.08342 0.96060 1.07959
50 2.20892 0.20892 0.53682 0.95720 2.64870
(2, 2.5) 0.02597 0.00050 0.00019 1.69865 0.07528 0.08869
2.62590 0.12590 0.20986 0.96520 1.67577
3.04545 0.29545 0.95233 0.94500 3.47825
(2.75, 3) 0.01512 0.00041 0.00009 2.04146 0.10388 0.13885
3.17595 0.17595 0.35139 0.94290 2.12751
1.05786 0.05786 0.07614 0.94250 1.07507
(1, 1.75) 0.06308 -0.00007 0.00032 1.15112 0.02341 0.01832
1.78736 0.03736 0.03713 0.96820 0.74201
100 2.08837 0.08837 0.21655 0.96210 1.76962
(2, 2.5) 0.02591 0.00044 0.00009 1.65477 0.03139 0.03758
2.55273 0.05273 0.08664 0.96840 1.12975
2.88369 0.13369 0.39611 0.95400 2.36377
(2.75, 3) 0.01500 0.00028 0.00004 1.98458 0.04700 0.05960
3.07958 0.07958 0.14831 0.94680 1.45343

TABLE 2
Descriptive statistics for the considered data sets.

Data Size Mean Median Mode Skewness Kurtosis variance IQR

Data set 1 100 2.6215 2.7000 2.7500 0.3681 0.1055 1.0278 1.3800
Data set 2 20 1.9000 1.7000 1.7500 1.7198 2.9241 0.4958 0.5750
264 A.S. Yadav, S. Shukla, N. Jaiswal, S.K. Singh and D. Koley

Boxplot of data set 1 Histogram of data set 1

WXGED

0.4
5

WED
GLD

Relative Frequency
GED
XGED
4
Observations

0.3
RD
ED
LD
3

XGD

0.2
2

0.1
1

0.0
0 1 2 3 4 5 6
Observations

Figure 5 – Boxplot and Histogram with theoretical density for Data set 1.

model is that the generative distribution of our proposed model is Exponential. The
comparison among the distributions is done using the Akaike information criterion
(AIC), Bayesian information criterion (BIC), and Kolmogorov–Smirnov (K-S) distance
values. The distributions are assigned a rank based on the smallest to largest K-S values.
From Table 3, we can see that our proposed model provides the best fit among all the
other distributions. Similarly, for Data set 2, we have compared the WXGED model

TABLE 3
Model fitting summary of Data set 1.

Model Estimates -LL AIC BIC K-S Rank

WXGED υ = 3.43523, ζ = 2.39482 142.65490 289.30990 294.52020 0.08965 1st
WED α1 = 0.00774, λ = 0.75912 166.37800 336.75590 341.96630 0.20661 5th
GLD α2 = 1.24695, θ1 = 5.87928 144.92840 293.85680 299.06710 0.10118 2nd
GED α3 = 7.78938, β = 1.01317 146.18060 296.36120 301.57160 0.10768 3rd
RD θ2 = 1.98617 149.50040 301.00090 303.60600 0.13832 4th
XGED θ3 = 0.90186 186.89680 375.79370 378.39880 0.33383 9th
LD θ4 = 0.61732 181.75620 365.51240 368.11750 0.26326 6th
ED θ5 = 0.38146 196.37470 394.74930 397.35452 0.32058 8th
XGD θ6 = 0.85092 184.65540 373.31080 378.52110 0.29058 7th

with WED, weighted Maxwell distribution (WMD), RD, XGED, LD, ED, and XGD.
The comparisons are made based on AIC, BIC, and K-S values. From Table 4, we can see
that our proposed model provides the best fit among all the other distributions. TTT
plot for the Data set 2 and eCDF vs theoretical CDF plots are given in Figure 8. Ta-
ble 5 contains the ML estimators of the parameters, survival, and the hazard function
Development and Estimation of Weighted Xgamma Exponential Distribution 265

1.0
1.0

0.8
0.8

0.6
0.6

ecdf
eCDF
T(i/n)

WXGED
WED

0.4
0.4

GLD
GED
XGED

0.2
0.2

RD
ED
LD
0.0 XGD
0.0

0.0 0.2 0.4 0.6 0.8 1.0 0 2 4 6 8 10

i/n Observations

Figure 6 – TTT plot and eCDF vs Theoretical CDF plot for Data set 1.

Boxplot of data set 2 Histogram of data set 2

1.0
1.0 1.5 2.0 2.5 3.0 3.5 4.0

WXGED
WED
WMD
0.8
Relative Frequency

RD
XGED
Observations

ED
0.6

LD
XGD
0.4
0.2
0.0

1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5

Observations

Figure 7 – Boxplot and Histogram with theoretical density for Data set 2.
266 A.S. Yadav, S. Shukla, N. Jaiswal, S.K. Singh and D. Koley

TABLE 4
Model fitting summary of Data set 2.

Model Estimates -LL AIC BIC K-S Rank

WXGED υ = 6.91655, ζ = 5.19197 17.85488 39.70976 41.70122 0.17416 1st
WED α1 = 0.01439, λ1 = 1.04551 26.16318 56.32636 58.31782 0.32212 4th
WMD α2 = 1.15002, λ2 = 1.69344 19.17005 42.34010 44.33156 0.19672 2nd
RD λ4 = 1.42846 22.47881 46.95762 47.95335 0.25658 3rd
XGED λ5 = 1.27147 30.90790 63.81580 64.81153 0.45961 8th
LD λ6 = 0.81612 30.24955 62.49910 63.49483 0.39108 5th
ED λ7 = 0.52632 32.83708 67.67416 68.66989 0.43951 6th
XGD λ8 = 1.10748 31.50824 67.01649 69.00795 0.42915 7th
1.0
1.0

0.8
0.8

0.6
0.6

ecdf

eCDF
T(i/n)

WXGED
WED
0.4
0.4

WMD
RD
XGED
0.2
0.2

ED
LD
XGD
0.0
0.0

0.0 0.2 0.4 0.6 0.8 1.0 1.0 1.5 2.0 2.5 3.0 3.5 4.0
i/n Observations

Figure 8 – TTT plot and eCDF vs Theoretical CDF plot for Data set 2.
Development and Estimation of Weighted Xgamma Exponential Distribution 267

(evaluated at the median of the corresponding data sets) and respective 95% ACI of dis-
tribution parameters υ and ζ . Taking one data set with a larger sample size and the other
one with a smaller sample size is motivated by the desire to see how the proposed esti-
mators for WXGED perform on both smaller and larger data sets. The goodness-of-fit
of the distributions is solely assessed through the alignments of the theoretical CDF and
the eCDF plots here, as the goodness-of-fit tests might perform poorly on small data
sets.

TABLE 5
ML estimates of parameters, survival/reliability characteristics, and 95% ACIs of parameters.

95% ACI
MLE υ ζ
υ̂ ζˆ Lower Upper Lower Upper Ŝ(t )| t =median ĥ(t )| t =median
Data set 1 3.43523 2.39482 1.65924 5.21123 1.67101 3.11863 0.42100 0.85031
Data set 2 6.91655 5.19197 3.14734 10.68577 3.05386 7.33009 0.59241 1.15648

6. CONCLUDING REMARKS

In conclusion, this study presents a newly developed weighted version of the XGED,
motivated by the broad range of applications that weighted distributions offer across
various fields. The newly introduced weighted distribution has been thoroughly inves-
tigated in terms of its statistical properties, with a specific focus on its length-biased and
area-biased forms. One of the key findings is that the proposed distribution exhibits
an increasing failure rate property, which makes it particularly well-suited for modeling
survival and reliability data in practical applications. The unknown parameters of the
distribution, along with its survival/reliability characteristics, have been estimated us-
ing the ML estimation method. Additionally, ACIs based on the ML estimation method
have been proposed for different model parameter variations, and their corresponding
CP and AW have been reported. A Monte Carlo simulation study has been conducted
to evaluate the performance of the estimators, comparing them in terms of their MSE.
To validate the practical applicability of the proposed weighted distribution, two real-
world data sets have been analyzed. The findings demonstrate that the proposed distri-
bution provides a superior fit compared to several existing weighted distributions and
other commonly used distributions with increasing failure rates. This highlights the
potential of the newly introduced distribution in real-life data modeling, particularly in
fields where reliability and survival data are of interest. All computational work for this
study has been performed using R software (version 4.2.1). Overall, the study not only
introduces a versatile new weighted distribution but also establishes its practical utility
and improved performance through a comprehensive analysis of theoretical properties,
simulation studies, and real-world applications.
268 A.S. Yadav, S. Shukla, N. Jaiswal, S.K. Singh and D. Koley

ACKNOWLEDGEMENTS

The authors express their gratitude to the editor and anonymous referees for their con-
structive feedback, which significantly improved the quality of this article. The first
author sincerely acknowledges Banaras Hindu University, Varanasi India, for providing
financial support under the Institute of Eminence Scheme (Dev. 6031) to carry out this
research work.

REFERENCES

R. A. FISHER (1934). The effect of methods of ascertainment upon the estimation of fre-
quencies. Annals of Eugenics, 6, no. 1, pp. 13–25.

R. E. GLASER (1980). Bathtub and related failure rate characterizations. Journal of the
American Statistical Association, 75, no. 371, pp. 667–672.

L. R. GROSENBAUGH (1958). Point sampling and line sampling: probability theory, ge-
ometric implications, synthesis, vol. 160. Southern Forest Experiment Station, Forest
Service, US Department of Agriculture.

A. J. GROSS, V. CLARK (1975). Survival Distributions: Reliability Applications in the

Biomedical Sciences. John Wiley & Sons, New York.

R. C. GUPTA, J. P. KEATING (1986). Relations for reliability measures under length biased
sampling. Scandinavian Journal of Statistics, pp. 49–56.

K. JAIN, H. SINGH, I. BAGAI (1989). Relations for reliability measures of weighted dis-
tributions. Communications in Statistics-Theory and Methods, 18, no. 12, pp. 4393–
4412.

D. T. L AROSE, D. K. DEY (1998). Modeling publication bias using weighted distributions

in a Bayesian framework. Computational Statistics & Data Analysis, 26, no. 3, pp.
279–302.

S. R. LELE, J. L. KEIM (2006). Weighted distributions and estimation of resource selection

probability functions. Ecology, 87, no. 12, pp. 3021–3028.

A. K. NANDA, S. BHATTACHARJEE, S. ALAM (2007). Properties of aging intensity func-

tion. Statistics & Probability Letters, 77, no. 4, pp. 365–373.

M. D. NICHOLS, W. J. PADGETT (2006). A bootstrap control chart for weibull percentiles.

Quality and Reliability Engineering International, 22, no. 2, pp. 141–151.

G. P. PATIL, J. ORD (1976). On size-biased sampling and related form-invariant weighted

distributions. Sankhyā: The Indian Journal of Statistics, Series B, pp. 48–61.
Development and Estimation of Weighted Xgamma Exponential Distribution 269

G. P. PATIL, C. R. RAO (1978). Weighted distributions and size-biased sampling with

applications to wildlife populations and human families. Biometrics, pp. 179–189.
C. R. RAO (1965). On discrete distributions arising out of methods of ascertainment.
Sankhyā: The Indian Journal of Statistics, Series A, pp. 311–324.
M. SHAKED, J. G. SHANTHIKUMAR (1994). Stochastic Orders and their Applications.
Academic Press, New York.
K. SHUKLA (2019). A comparative study of one parameter lifetime distributions. Biomet-
rics & Biostatistics International Journal, 8, no. 4, pp. 111–123.
S. SUNOJ, S. MAYA (2006). Some properties of weighted distributions in the context of
repairable systems. Communications in Statistics—Theory and Methods, 35, no. 2,
pp. 223–228.
W. WARREN, P. OLSEN (1964). A line intersect technique for assessing logging waste.
Forest Science, 10, no. 3, pp. 267–276.
A. S. YADAV, S. SHUKLA, H. GOUAL, M. SAHA, H. M. YOUSOF (2022). Validation of
Xgamma exponential model via Nikulin-Rao-Robson goodness-of-fit-test under complete
and censored sample with different methods of estimation. Statistics, Optimization &
Information Computing, 10, no. 2, pp. 457–483.