Received: 19 July 2022 Revised: 8 April 2023 Accepted: 17 April 2023

DOI: 10.1002/qre.3362

RESEARCH ARTICLE

Statistical inference for multi stress–strength reliability

based on progressive first failure with lifetime inverse
Lomax distribution and analysis of transformer
insulation data

Dina A. Ramadan1 Ehab M. Almetwally2 Ahlam H. Tolba1

1 Faculty
of Science, Department of
Mathematics, Mansoura University, Abstract
Mansoura, Egypt This research aims to develop a multireliability inference method for stress–
2 Facultyof Business Administration, strength variables based on progressive first failure and an inverse Lomax
Department of Statistics, Delta University
for Science and Technology, Belkas, Egypt
distribution. This research examines the difficulties associated with estimating
the stress–strength reliability function, 𝑅, when 𝑋, 𝑌, and 𝑍 are drawn from three
Correspondence different Inverse Lomax distributions. On the basis of progressive first-failure
Dina A. Ramadan.
Email: [email protected] censored samples, reliability estimators for multi-stress–strength Inverse Lomax
distributions are estimated using the maximum likelihood, maximum product of
spacing, and Bayesian estimation methods. The Bayes estimate of 𝑅 is obtained
using the MCMC method for a symmetric loss function. Monte Carlo simula-
tions and real-data applications are used to assess and compare the performance
of the various suggested estimators.

KEYWORDS
bootstrap confidence intervals, loss functions, Monte Carlo Markov chain, multi-stress–
strength, progressive first failure censoring

1 INTRODUCTION

Reliability of technical equipment in the power industry. It is, therefore, necessary, in the absence of scientific and research
potential in a distribution organizations (e.g., power plants, heating plants), to achieve the objectives of the proposed
activities, that is, in-depth analysis of undesirable impacts on the state of devices, design of measurements and their
verification, and design of new diagnostic procedures for improving reliability of power transformers. Unless we want
to determine real insulating state of transformer and then lifetime of insulation is necessary to analyze some measure-
ments in individual type of assays and then determine their exactness and reliability with mathematical model. We can
exactly prove the importance of these assays by mathematical and statistical models in region of analyze of insulating
state of transformers, see Zhou et al.1 When systems fail to perform in their extreme working settings, it is a common
occurrence in real-life scenarios. When systems cross their lower, upper, or both extreme working environments, they
fail to perform their intended functions. In every area of engineering, there is a growing need to create new methodolo-
gies for estimating the reliability of technological goods. This is caused by a combination of two main causes: the first is
the rapid expansion of liberalization (which is discussed below with reference to the power systems); the second is the
constant level of technological innovation, which results in components with ever-increasing reliability values, implying
a lack of failure data. Additionally, the propensity for very quick product development times and constrained budgets

Qual Reliab Engng Int. 2023;1–24. wileyonlinelibrary.com/journal/qre © 2023 John Wiley & Sons Ltd. 1
 10991638, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/qre.3362 by Central University Of Rajasthan, Wiley Online Library on [04/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2 RAMADAN et al.

necessitate that reliability tests be performed under strict time constraints, which frequently results in no failures during
such tests.
The value of the probability 𝑅 = 𝑃(𝑋 < 𝑌), also known as stress–strength reliability, has been extensively researched
in the literature. When applied stress exceeds the system’s strength, a system working under such a stress–strength
setup ceases to function. Several authors assuming various lifetime distributions for the stress–strength random
variate.2-10
The stress–strength model 𝑅 = 𝑃(𝑋 < 𝑌) was used, and the study was expanded to include multicomponent systems.
However, an equally important practical case in which objects perform in both extreme lower and extreme upper working
contexts receives less attention. If electrical equipment is placed below or over a particular power supply, for example, it
will fail. There are many of these applications, and they are all basic and natural, reflecting sound relationships between
diverse real-world occurrences. Strength 𝑌 should not only be greater than stress 𝑋, but also lower than stress 𝑍, accord-
ing to this useful relationship in genetics and psychology. Many academics have looked at estimating the stress–strength
parameter for a variety of statistical models. The estimation of 𝑅 = 𝑃(𝑋 < 𝑌 < 𝑍) using independent samples was inves-
tigated by Chandra and Owen,11 Yousef and Almetwally,12 Yousef and Almetwally,13 Dutta and Sriwastav,14 and Ivshin.15
Furthermore, the study of stress–strength models has been expanded to multicomponent systems, also known as multi-
component systems. Despite the fact that Bhattacharyya and Johnson16 developed the multicomponent stress–strength
model, it has garnered a lot of attention in recent years and has been explored by numerous researchers for both complete
and censored data. Some of the recently appeared articles include Tolba et al.,17 Ramadan et al.,18 Kotb and Raqab,19 Mau-
rya and Tripathi,20 Mahto et al.,21 Wang et al.,22 Jha et al.,23 El-Sagheer et al.,24 Rasekhi et al.,25 Alotaibi et al.,26 Maurya
et al.,27 Sarhan et al.,28 Sarhan et al.,29 Kohansal and Shoaee,30 Muse et al.,31 Jana and Bera,32 Ahmad et al.,33 Ahmad
et al.34 , Ramadan35 and Wu and Kuş.36
Well-known censoring techniques include the type-II censoring scheme, progressive type-II censoring scheme, and
progressive first failure censoring system. As a new life-test plan, Wu and Kuş36 presented the progressive first failure
censoring scheme, which combines progressive type-II censoring and first failure censoring strategies. The following is
a summary of the progressive first failure censoring system: Assume that a life-test is given to 𝑛 different groups, each
containing 𝑘 different objects. The 𝑅1 units and the group in which the first failure is detected are randomly eliminated
from the experiment after the first failure 𝑌1;𝑚,𝑛,𝑘 .
At the time of the second failure 𝑌2;𝑚,𝑛,𝑘 , the 𝑅2 units and the group in which the second failure is observed are randomly
withdrawn from the remaining live (𝑛 − 𝑅1 − 2) groups. At the end, when the mth observation 𝑌𝑚;𝑚,𝑛,𝑘 fails, the rest of
the live units 𝑅𝑚 , (𝑚 < 𝑛) are withdrawn from the test. Then, the obtained ordered observations 𝑌1;𝑚,𝑛,𝑘 < ⋯ <𝑌𝑚;𝑚,𝑛,𝑘
are called progressively first-failure censored with progressive censored scheme specified by 𝑅 = (𝑅1 , 𝑅2 , … , 𝑅𝑚 ), where
∑𝑚
𝑚 failures and sum of all removals sums to 𝑛, that is, 𝑛 = 𝑚 + 𝑖=1 𝑅𝑖 . One may notice that a special case with
𝑅1 = 𝑅2 = ⋯ = 𝑅𝑚 = 0 reduces the progressive first-failure censoring to first-failure censoring scheme. Similarly, with
𝑅1 = 𝑅2 = ⋯ = 𝑅𝑚−1 = 0 and 𝑅𝑚 = 𝑛 − 𝑚, first-failure type-II censoring comes as a particular case of this censoring
scheme. With the assumption that each group contains exactly one unit, that is, 𝑘 = 1, the progressive first-failure censor-
ing scheme reduces to the progressive type-II censoring scheme, Elnagar et al.,37 Khalifa et al.38 and Tolba et al.39 Thus, a
generalization of progressive censoring is progressive first-failure censoring.
The inverse Lomax is an example of a second-type generalized beta distribution. In statistical applications, this is one of
the most important lifetime models. According to Kleiber and Kotz,40 the inverse Lomax distribution has been studied in
a range of domains, including stochastic modelling, economics, statistics sciences, reliability analysis, and life-testing.41
The Inverse Lomax distribution was used by Ziegel42 to get Lorenz to construct relationships between sorted data. In
McKenzie et al.,43 this lifetime distribution was applied to geophysical data, specifically the diameters of ground fibers
in California.
Let 𝑌 is a random variable of the Lomax distribution, then the inverse Lomax distribution of 𝑋 = 1∕𝑌 is also invisible.
The cumulative distribution function (cdf), probability density function (pdf), the survival rate function, and the hazard
rate function of Inverse Lomax distribution IL(𝛼, 𝛽) are given, respectively, by

𝐹(𝑥; 𝛼, 𝛽) = (1 + 𝛽𝑥 −1 )−𝛼 ; 𝛼, 𝛽 > 0, 𝑥 > 0, (1.1)

𝑓(𝑥; 𝛼, 𝛽) = 𝛼𝛽𝑥 −2 (1 + 𝛽𝑥 −1 )−(1+𝛼) ; 𝛼, 𝛽 > 0, 𝑥 > 0, (1.2)

𝑆(𝑥; 𝛼, 𝛽) = 1 − (1 + 𝛽𝑥 −1 )−𝛼 ; 𝛼, 𝛽 > 0, 𝑥 > 0 (1.3)

 10991638, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/qre.3362 by Central University Of Rajasthan, Wiley Online Library on [04/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAMADAN et al. 3

and

𝛼𝛽𝑥 −2 (1 + 𝛽𝑥 −1 )−(1+𝛼)
ℎ(𝑥) = ; 𝛼, 𝛽 > 0, 𝑥 > 0, (1.4)
1 − (1 + 𝛽𝑥 −1 )−𝛼

where 𝛼 is shape parameter and 𝛽 is the scale parameter.

As a result, multicomponent reliability inference is required, with stress–strength variables following an IL distribution
based on progressive first-failure. The problem of predicting the stress–strength function 𝑅, where 𝑋, 𝑌, and 𝑍 are three
independent IL, is addressed in this paper. Also discussed are asymptotic confidence intervals, boot-p and boot-t meth-
ods, as well as likelihood estimates based on progressive first-failure censored for point estimation. Bayesian estimate
methods based on progressive first-failure censored are created using Markov chain Monte Carlo (MCMC). The symmet-
ric loss function was used for Bayesian estimation. The reliability of multi-stress–strength IL distribution was assessed
using loss functions based on progressive first-failure censored samples. Using Monte Carlo simulations and real data, the
performance of the various proposed estimators is tested and compared.
The following are contents for this paper. The maximum likelihood and maximum product of spacing for point and
interval estimates are developed in Section 2. Bootstrap confidence interval, boot-p, and boot-t are discussed in Section 3.
Section 4 discusses Bayesian estimation methods, including MCMC techniques. The Bayes estimate of R is presented in
this section. Additional simulation studies are included in Section 5. A real data application is demonstrated in Section 6.
Finally, Section 7 introduces the paper’s conclusion.

2 CLASSICAL INFERENCE METHODS

The traditional point and interval estimating methodologies are covered in this section, including maximum likelihood
estimation for R point estimates and asymptotic, boot-p, and boot-t intervals for R interval estimates.

2.1 Maximum likelihood inference of R

Let 𝑋 ∼ 𝐼𝐿(𝛼1 , 𝛽), 𝑌 ∼ 𝐼𝐿(𝛼2 , 𝛽), 𝑍 ∼ 𝐼𝐿(𝛼3 , 𝛽) and they are independent. Assuming that 𝛽 is known, we have
∞ ∞
𝑅 = 𝑃(𝑌 < 𝑋 < 𝑍) = 𝐹𝑥 (𝑦)𝑑𝐹𝑌 (𝑥) − 𝐹𝑋 (𝑦)𝐹𝑍 (𝑦)𝑑𝐹𝑌 (𝑦), (2.1)
∫−∞ ∫−∞

𝛼2 𝛼2
= − (2.2)
𝛼1 + 𝛼2 𝛼1 + 𝛼2 + 𝛼3

𝛼2 𝛼3
= . (2.3)
(𝛼1 + 𝛼2 )(𝛼1 + 𝛼2 + 𝛼3 )

2.2 Maximum likelihood of R under progressively first failure

To get the MLE of 𝑅, we first need to get the MLEs of 𝛼1 , 𝛼2 , and 𝛼3 .

Let (𝑋1;𝑚1 ,𝑛1 ,𝑘1 , 𝑋2;𝑚1 ,𝑛1 ,𝑘1 , … , 𝑋𝑛1 ;𝑚1 ,𝑛1 ,𝑘1 ), (𝑌1;𝑚2 ,𝑛2 ,𝑘2 , 𝑌2;𝑚2 ,𝑛2 ,𝑘2 , … , 𝑌𝑛2 ;𝑚2 ,𝑛2 ,𝑘2 ) and
(𝑍1;𝑚3 ,𝑛3 ,𝑘3 , 𝑍2;𝑚3 ,𝑛3 ,𝑘3 , … , 𝑍𝑛3 ;𝑚3 ,𝑛3 ,𝑘3 ), be three progressively first failure censored samples from IL (𝛼𝑖 , 𝛽 ), 𝑖 = 1, 2, 3
distribution with censoring schemes 𝑅𝑥 = (𝑅𝑥1 , 𝑅𝑥2 , … , 𝑅𝑥𝑚1 ), 𝑅𝑦 = (𝑅𝑦1 , 𝑅𝑦2 , … , 𝑅𝑦𝑚2 ), and 𝑅𝑧 = (𝑅𝑧1 , 𝑅𝑧2 , … , 𝑅𝑧𝑚3 ). The
likelihood function is based on Balakrishnan et al.44 and Wu and Kuş.36
Therefore, using the expressions from Equations (2.1) and (2.2), the likelihood function of 𝛼1 , 𝛼2 , and 𝛼3 is given by

3 𝑚1 ( )−(1+𝛼1 ) [ ( )−𝛼1 ]𝑘1 (𝑅𝑥𝑖 +1)−1

∏ ∏ 𝛽 𝛽
𝐿(𝛼1 , 𝛼2 , 𝛼3 , 𝛽|𝑑𝑎𝑡𝑎) = (𝑘𝑗 𝛼𝑗 )𝑚𝑗 𝛽 𝑚1 +𝑚2 +𝑚3 𝑥𝑖−2 1+ 1− 1+
𝑗=1 𝑖=1
𝑥𝑖 𝑥𝑖
 10991638, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/qre.3362 by Central University Of Rajasthan, Wiley Online Library on [04/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
4 RAMADAN et al.

𝑚2 ( )−(1+𝛼2 ) [ ( )−𝛼2 ]𝑘2 (𝑅𝑦𝑗 +1)−1

∏ 𝛽 𝛽
. 𝑦𝑗−2 1+ 1− 1+
𝑗=1
𝑦𝑗 𝑦𝑗

𝑚3 ( )−(1+𝛼3 ) [ ( )−𝛼3 ]𝑘3 (𝑅𝑧𝑙 +1)−1

∏ 𝛽 𝛽
. 𝑧𝑙−2 1+ 1− 1+ . (2.4)
𝑧𝑙 𝑧𝑙
𝑙=1

For the simplicity of notation, we will use 𝑥𝑖 instead of 𝑋𝑖;𝑚1 ,𝑛1 ,𝑘1 , similarity for 𝑦𝑗 and 𝑧𝑙 . The log-likelihood function can
be expressed as

3
[𝑚 𝑚2 𝑚3
]
∑ ∑1 ∑ ∑
𝓁(𝛼1 , 𝛼2 , 𝛼3 , 𝛽|𝑑𝑎𝑡𝑎) = (log 𝑘𝑗 + log 𝛼𝑗 ) + (𝑚1 + 𝑚2 + 𝑚3 ) log 𝛽 − 2 𝑥𝑖 + 𝑦𝑗 + 𝑧𝑙
𝑗=1 𝑖=1 𝑗=1 𝑙=1

𝑚1
∑ 𝑚2
∑
− (𝛼1 + 1) log(1 + 𝛽𝑥𝑖−1 ) − (𝛼2 + 1) log(1 + 𝛽𝑦𝑗−1 )
𝑖=1 𝑗=1

𝑚3 𝑚1
∑ ∑
− (𝛼3 + 1) log(1 + 𝛽𝑧𝑙−1 ) + (𝑘1 (𝑅𝑥𝑖 + 1) − 1) log(1 − (1 + 𝛽𝑥𝑖−1 )−𝛼1 )
𝑙=1 𝑖=1
𝑚2
∑
+ (𝑘2 (𝑅𝑦𝑗 + 1) − 1) log(1 − (1 + 𝛽𝑦𝑗−1 )−𝛼2 )
𝑗=1

𝑚3
∑
+ (𝑘3 (𝑅𝑧𝑙 + 1) − 1) log(1 − (1 + 𝛽𝑧𝑙−1 )−𝛼3 ). (2.5)
𝑙=1

The inverse Lomax distribution parameters 𝛼1 , 𝛼2 , 𝛼3 and 𝛽 and the solution with the maximum likelihood estima-
tor under multi-stress–strength reliability is established. By differentiating the natural logarithm, 𝓁(𝛼1 , 𝛼2 , 𝛼3 , 𝛽|𝑑𝑎𝑡𝑎),
partially relative to 𝛼1 , 𝛼2 , 𝛼3 and 𝛽 then equivalent to zero.
( )
𝛽
𝑚1 ( ) 𝑚1 log 1 +
𝜕𝓁 𝑚 ∑ 𝛽 ∑ 𝑥𝑖
= 1− log 1 + − (𝑘1 (𝑅𝑥𝑖 + 1) − 1)(1 + 𝛽𝑥𝑖−1 )−𝛼1 , (2.6)
𝜕𝛼1 𝛼1 𝑖=1
𝑥𝑖 𝑖=1
−1 −𝛼1
1 − (1 + 𝛽𝑥 ) 𝑖
( )
𝛽
𝑚2 ( ) 𝑚2 log 1 +
𝜕𝓁 𝑚 ∑ 𝛽 ∑ 𝑦𝑗
= 2− log 1 + − (𝑘2 (𝑅𝑦𝑗 + 1) − 1)(1 + 𝛽𝑦𝑗−1 )−𝛼2 (2.7)
𝜕𝛼2 𝛼2 𝑗=1 𝑦𝑗 𝑗=1
−1
1 − (1 + 𝛽𝑦 )−𝛼2 𝑗

and
( )
𝛽
𝑚3 ( ) 𝑚3 log 1 +
𝜕𝓁 𝑚 ∑ 𝛽 ∑ 𝑧𝑙
= 3− log 1 + − (𝑘3 (𝑅𝑧𝑙 + 1) − 1)(1 + 𝛽𝑧𝑙−1 )−𝛼3 . (2.8)
𝜕𝛼3 𝛼3 𝑧𝑙 −1 −𝛼3
1 − (1 + 𝛽𝑧 )
𝑙=1 𝑙=1 𝑙

𝜕𝓁 𝜕𝓁 𝜕𝓁
The system of normal equations = 0, = 0, and = 0 has not closed form of its solution, so the numerical
𝜕𝛼1 𝜕𝛼2 𝜕𝛼3
techniques to estimate the unknown parameters 𝛼1 , 𝛼2 , and 𝛼3 are used.
The Delta method can be used to estimate the IL distribution’s multi-stress–strength reliability function 𝑅, as shown
below.

𝛼̂2 𝛼̂3
𝑅̂ = . (2.9)
(𝛼̂1 + 𝛼̂2 )(𝛼̂1 + 𝛼̂2 + 𝛼̂3 )
 10991638, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/qre.3362 by Central University Of Rajasthan, Wiley Online Library on [04/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAMADAN et al. 5

2.3 Asymptotic confidence intervals

The 100(1 − 𝛾)% CIs for 𝛼1 , 𝛼2 , and 𝛼3 can be obtained from the usual asymptotic normality of the maximum likeli-
hood estimators with Var(𝛼̂1 𝑀𝐿 ), Var(𝛼̂2 𝑀𝐿 ) and Var(𝛼̂3 𝑀𝐿 ) estimated from the inverse of the observed Fisher information
matrix, that is, the inverse of the matrix of second derivatives of the log-likelihood function locally at 𝛼̂1 𝑀𝐿 , 𝛼̂2 𝑀𝐿 and
𝛼̂3 𝑀𝐿 .
From the log-likelihood function in Equation (2.5), the second derivatives with respect to 𝛼1 , 𝛼2 , and 𝛼3 are given by

( ( ))2
𝛽
𝑚1
∑ log 1 +
𝜕2 𝓁 −𝑚 𝑥𝑖
2
= − 21 − (𝑘1 (𝑅𝑥𝑖 + 1) − 1)(1 + 𝛽𝑥𝑖−1 )−𝛼1 ( )2 ,
𝜕𝛼1 𝛼1 𝑖=1 1 − (1 + 𝛽𝑥 −1 )−𝛼1 𝑖
( ( ))2
𝛽
𝑚 log 1 +
𝜕2 𝓁 −𝑚2 ∑ 2 𝑦𝑗
2
= − 2
− (𝑘2 (𝑅𝑦𝑗 + 1) − 1)(1 + 𝛽𝑦𝑗−1 )−𝛼2 ( )2
𝜕𝛼2 𝛼2 𝑗=1 1 − (1 + 𝛽𝑦𝑗−1 )−𝛼2

and
( ( ))2
𝛽
∑ 𝑚3 log 1 +
𝜕2 𝓁 −𝑚 𝑧𝑙
2
= − 23 − (𝑘3 (𝑅𝑧𝑙 + 1) − 1)(1 + 𝛽𝑧𝑙−1 )−𝛼3 ( )2 .
𝜕𝛼3 𝛼3 𝑙=1 −1
1 − (1 + 𝛽𝑧 )−𝛼3 𝑙

Therefore, the observed Fisher information matrix 𝐼̂𝑖𝑗 = 𝐸[−𝜕 2 𝓁∕𝜕𝜙𝑖 𝜕𝜙𝑗 ], where 𝑖, 𝑗 = 1, 2, 3 and 𝜙 = (𝜙1 , 𝜙2 , 𝜙3 ) =
(𝛼1 , 𝛼2 , 𝛼3 ), which is obtained by dropping the expectation operator 𝐸, will be used to construct confidence intervals for
the parameters, see Lawless.45 Hence, the observed information matrix is given by

2 𝜕2 𝓁 𝜕2 𝓁 ⎞
⎛ −𝜕 𝓁 − −
⎜ 𝜕𝛼12 𝜕𝛼1 𝜕𝛼2 𝜕𝛼1 𝜕𝛼3 ⎟
⎜ ⎟
⎜ 𝜕2 𝓁 𝜕2 𝓁 𝜕2 𝓁 ⎟
̂ 1 , 𝛼2 , 𝛼3 ) = ⎜−
𝐼(𝛼 − − ⎟.
⎜ 𝜕𝛼2 𝜕𝛼1 𝜕𝛼22 𝜕𝛼2 𝜕𝛼3 ⎟
⎜ ⎟
⎜ 𝜕2 𝓁 𝜕2 𝓁 𝜕2 𝓁 ⎟
⎜− − − 2 ⎟
⎝ 𝜕𝛼3 𝜕𝛼1 𝜕𝛼3 𝜕𝛼2 𝜕𝛼3 ⎠

Therefore, inverting the observed information matrix 𝐼(𝛼 ̂ 1 , 𝛼2 , 𝛼3 ) is used to obtain the asymptotic variance–covariance
̂ −1
matrix for the MLEs, where 𝐼 (𝛼1 , 𝛼2 , 𝛼3 ) is obtained by

⎛ 𝑣𝑎𝑟(𝛼
ˆ1 ) 𝑐𝑜𝑣(𝛼1 , 𝛼2 ) 𝑐𝑜𝑣(𝛼1 , 𝛼3 )⎞
⎜ ⎟
𝐼̂ −1 (𝛼1 , 𝛼2 , 𝛼3 ) = ⎜𝑐𝑜𝑣(𝛼2 , 𝛼1 ) ˆ2 )
𝑣𝑎𝑟(𝛼 𝑐𝑜𝑣(𝛼2 , 𝛼3 )⎟.
⎜ ⎟
⎜𝑐𝑜𝑣(𝛼 , 𝛼 ) 𝑐𝑜𝑣(𝛼 , 𝛼 ) ˆ
𝑣𝑎𝑟(𝛼 ) ⎟
⎝ 3 1 3 2 3 ⎠

Thus, the 100(1 − 𝛾)% normal approximate CIs for (𝛼1 , 𝛼2 , 𝛼3 ) are
√ √ √
ˆ1 ± 𝑍 𝛾 𝑣𝑎𝑟(𝛼1 ), 𝛼
𝛼 ˆ ˆ
ˆ2 ± 𝑍 𝛾 𝑣𝑎𝑟(𝛼2 ) and 𝛼 ˆ3 ),
ˆ3 ± 𝑍 𝛾 𝑣𝑎𝑟(𝛼 (2.10)
2 2 2

𝛾
where 𝑍 𝛾 is the percentile of the standard normal distribution with right-tail probability .
2 2
In order to compute the asymptotic CI of the multi-stress–strength reliability function, we must first find the variance. To
find approximate estimates of 𝑅̂ variance, the delta approach was applied. For MLE functions, the delta technique gives
a general method for determining CIs. It takes a function that is too complicated to calculate the variance analytically,
 10991638, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/qre.3362 by Central University Of Rajasthan, Wiley Online Library on [04/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
6 RAMADAN et al.

makes a linear approximation of it, and then computes the variance of the smaller linear function that may be used for
large sample inference, see Washington et al.46 We define
( )
𝜕𝑅 𝜕𝑅 𝜕𝑅
𝐺́ = , , ,
𝜕𝛼1 𝜕𝛼2 𝜕𝛼3

where
[ ]
𝜕𝑅 𝛼2 𝛼3 𝛼3 + 2(𝛼1 + 𝛼2 )2
= , (2.11)
𝜕𝛼1 (𝛼1 + 𝛼2 )2 (𝛼1 + 𝛼2 + 𝛼3 ))
2

𝜕𝑅 (𝛼 + 𝛼3 )(𝛼1 + 𝛼2 )2 − 𝛼1 (𝛼1 + 𝛼2 + 𝛼3 )2
= 1 (2.12)
𝜕𝛼2 (𝛼1 + 𝛼2 )2 (𝛼1 + 𝛼2 + 𝛼3 )2

and

𝜕𝑅 −𝛼2
= . (2.13)
𝜕𝛼3 (𝛼1 + 𝛼2 + 𝛼3 )2

̂ are given, respectively, by

Then, the approximate estimates of 𝑉𝑎𝑟(𝑅)
[ −1 ]
ˆ 𝑅)
𝑉𝑎𝑟( ́
̂ ≃ 𝐺𝐼 𝐺 (𝛼 .
1 ,𝛼2 ,𝛼3 )= (𝛼̂1 𝑀𝐿 ,𝛼̂2 𝑀𝐿 ,𝛼̂3 𝑀𝐿 )

In addition, calculate the following statistic

( )
𝑅̂ − 𝑅
√ ∼ 𝑁(0, 1),
ˆ 𝑅)
𝑉𝑎𝑟( ̂

these results yield the approximate CIs for 𝑅 as

√
𝑅̂ ± 𝑍 𝛾 ˆ 𝑅).
𝑉𝑎𝑟( ̂ (2.14)
2

2.4 Maximum product of spacing (MPS) estimation

For a stress–strength model, the maximum product spacing is defined as follows:

𝑚+1
∏ 𝑚
∏ 𝑘(𝑅𝑖 +1)−1
𝑆(𝛼1 , 𝛼2 , 𝛼3 , 𝛽|𝑑𝑎𝑡𝑎) = (𝐹(𝑡𝑖 ) − 𝐹(𝑡𝑖−1 )) (1 − 𝐹(𝑡𝑖 )) .
𝑖=1 𝑖=1

The geometric mean of the spacing is given by

1
(𝑚+1 ) 𝑚+1
∏
𝐺= 𝐷𝑖 , (2.15)
𝑖=1

where

𝐷1 = 𝐹(𝑡1 ),

𝐷2 = 𝐹(𝑡𝑖 ) − 𝐹(𝑡𝑖−1 ) = 𝐹(𝑡2∶𝑚 ), 𝑖 = 1, 2, … .., 𝑚,

𝐷𝑚 = 1 − 𝐹(𝑡𝑚 ). (2.16)
 10991638, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/qre.3362 by Central University Of Rajasthan, Wiley Online Library on [04/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAMADAN et al. 7

Then, the MPS based on a progressive first failure is obtained by

[ 𝑚
] 𝑚
∏ ∏ 𝑘(𝑅𝑖 +1)−1
𝐿𝑀𝑃𝑆 (𝛼1 , 𝛼2 , 𝛼3 , 𝛽|𝑑𝑎𝑡𝑎) = 𝐴𝑘𝑚 𝐹(𝑡1 )(1 − 𝐹(𝑡𝑚 )) (𝐹(𝑡𝑖 ) − 𝐹(𝑡𝑖−1 )) (1 − 𝐹(𝑡𝑖 )) . (2.17)
𝑖=2 𝑖=1

So, the MPS for three progressive first failure becomes

[ 𝑚1
] 𝑚1
∏ ∏ 𝑘(𝑅𝑥𝑖 +1)−1
𝐿𝑀𝑃𝑆 (𝛼1 , 𝛼2 , 𝛼3 , 𝛽|𝑑𝑎𝑡𝑎) = 𝐴 𝐹(𝑥1 )(1 − 𝐹(𝑥𝑚1 )) (𝐹(𝑥𝑖 ) − 𝐹(𝑥𝑖−1 )) (1 − 𝐹(𝑥𝑖 ))
𝑖=2 𝑖=1
[ 𝑚2
]𝑚
𝑚 𝑚
∏ ( ) ∏2
( )𝑘(𝑅𝑦𝑗 +1)−1
.𝑘2 2 𝑘3 3 𝐹(𝑦1 )(1 − 𝐹(𝑦𝑚2 )) 𝐹(𝑦𝑗 ) − 𝐹(𝑦𝑗−1 ) 1 − 𝐹(𝑦𝑗 )
𝑗=2 𝑗=1
[ 𝑚3
] 𝑚3
𝑚
∏ ∏ 𝑘(𝑅𝑧𝑙 +1)−1
.𝑘1 1 𝐹(𝑧1 )(1 − 𝐹(𝑧𝑚3 )) (𝐹(𝑧𝑙 ) − 𝐹(𝑧𝑙−1 )) (1 − 𝐹(𝑧𝑙 )) . (2.18)
𝑙=2 𝑙=1

Substituting Equation (2.1) in Equation (2.18), we get

[( )−𝛼1 ( ( )−𝛼1 ) 𝑚1 (( )−𝛼1 ( )−𝛼1 )]
𝛽 𝛽 ∏ 𝛽 𝛽
𝐿𝑀𝑃𝑆 (𝛼1 , 𝛼2 , 𝛼3 , 𝛽|𝑑𝑎𝑡𝑎) = 1+ 1− 1+ 1+ − 1+
𝑥1 𝑥1 𝑖=2
𝑥𝑖 𝑥𝑖−1

𝑚1
∏
𝑚 𝑚 𝑚 [ ]𝑘1 (𝑅𝑥𝑖 +1)−1
. 𝐴𝑘1 1 𝑘2 2 𝑘3 3 1 − (1 + 𝛽𝑥𝑖−1 )𝛼1
𝑖=1
[( )−𝛼2 ( ( )−𝛼2 ) 𝑚2 (( )−𝛼2 ( )−𝛼2 )]
𝛽 𝛽 ∏ 𝛽 𝛽
. 1+ 1− 1+ 1+ − 1+
𝑦1 𝑦1 𝑗=2
𝑦𝑗 𝑦𝑗−1

𝑚2 [
∏ ]𝑘2 (𝑅𝑦𝑗 +1)−1
. 1 − (1 + 𝛽𝑦𝑗−1 )𝛼2
𝑗=1
[( )−𝛼3 ( ( )−𝛼3 ) 𝑚3 (( )−𝛼3 ( )−𝛼1 )]
𝛽 𝛽 ∏ 𝛽 𝛽
. 1+ 1− 1+ 1+ − 1+
𝑧1 𝑧1 𝑧𝑙 𝑧𝑙−1
𝑙=2
𝑚3
∏[ ]𝑘3 (𝑅𝑧𝑙 +1)−1
. 1 − (1 + 𝛽𝑧𝑙−1 )𝛼3 . (2.19)
𝑙=1

By taking log function on both sides, we obtain

log 𝐿𝑀𝑃𝑆 (𝛼1 , 𝛼2 , 𝛼3 , 𝛽|𝑑𝑎𝑡𝑎) = 𝑚1 log 𝑘1 + 𝑚2 log 𝑘2 + 𝑚3 log 𝑘3 − 𝛼1 log(1 + 𝛽𝑥1−1 )

𝑚1
∑
+ log(1 − (1 + 𝛽𝑥1−1 )−𝛼1 ) + log((1 + 𝛽𝑥1−1 )−𝛼1 − (1 + 𝛽𝑥𝑖−1
−1 −𝛼1
) )
𝑖=2
𝑚1
∑
+ (𝑘1 (𝑅𝑥𝑖 + 1) − 1) log(1 − (1 + 𝛽𝑥𝑖−1 )−𝛼1 ) − 𝛼2 log(1 + 𝛽𝑦1−1 ) +
𝑖=1
𝑚2
∑
+ log(1 − (1 + 𝛽𝑦1−1 )−𝛼2 ) + log((1 + 𝛽𝑦1−1 )−𝛼2 − (1 + 𝛽𝑦𝑗−1
−1 −𝛼2
) )
𝑗=2

𝑚2
∑
+ (𝑘2 (𝑅𝑦𝑗 + 1) − 1) log(1 − (1 + 𝛽𝑦𝑗−1 )−𝛼2 ) − 𝛼3 log(1 + 𝛽𝑧1−1 ) + log(1 − (1 + 𝛽𝑧1−1 )−𝛼3 )
𝑗=1
 10991638, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/qre.3362 by Central University Of Rajasthan, Wiley Online Library on [04/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
8 RAMADAN et al.

𝑚3 𝑚3
∑ ∑
+ log((1 + 𝛽𝑧1−1 )−𝛼3 − (1 + 𝛽𝑧𝑙−1
−1 −𝛼3
) )+ (𝑘3 (𝑅𝑧𝑙 + 1) − 1) log(1 − (1 + 𝛽𝑧𝑙−1 )−𝛼3 ).
𝑙=2 𝑙=1

(2.20)

In order to derive the normal equations for the unknown parameters, Equation (2.20) for parameters 𝛼1 , 𝛼2 , and 𝛼3 is
partially differentiated and zero equated. As a solution to the following equations, the estimators can be derived as
( (
)−𝛼1 )
𝛽 𝛽
1+ log 1 +
𝜕 log 𝑙𝑀𝑃𝑆 𝑥1 𝑥1
= log(1 + 𝛽𝑥1−1 ) + ( )−𝛼1
𝜕𝛼1 𝛽
1− 1+
𝑥1
( )−𝛼1 ( ( ) )−𝛼1 ( )
𝛽 𝛽 𝛽 𝛽
∑ 1+
𝑚1
𝑥𝑖
log 1 +
𝑥𝑖
− 1+
𝑥𝑖−1
log 1 +
𝑥𝑖−1
+ ( )−𝛼1 ( )−𝛼1
𝛽 𝛽
𝑖=2 1+ − 1+
𝑥𝑖 𝑥𝑖−1

𝑚1
∑ (1 + 𝛽𝑥𝑖−1 ) log(1 + 𝛽𝑥𝑖−1 )
− (𝑘1 (𝑅𝑥𝑖 + 1) − 1) ( )−𝛼1 , (2.21)
𝛽
𝑖=1 1− 1+
𝑥𝑖
( )−𝛼2 ( )
𝛽 𝛽
1+ log 1 +
𝜕 log 𝑙𝑀𝑃𝑆 𝑦1 𝑦1
= log(1 + 𝛽𝑦1−1 ) + ( )−𝛼2
𝜕𝛼2 𝛽
1− 1+
𝑦1
( )−𝛼2 ( ) ( )−𝛼2 ( )
𝛽 𝛽 𝛽 𝛽
𝑚2 1+ log 1 + − 1+ log 1 +
∑ 𝑦𝑗 𝑦𝑗 𝑦𝑗−1 𝑦𝑗−1
+ ( )−𝛼2 ( )−𝛼2
𝑗=2 𝛽 𝛽
1+ − 1+
𝑦𝑗 𝑦𝑗−1

𝑚1
∑ (1 + 𝛽𝑦𝑗−1 ) log(1 + 𝛽𝑦𝑗−1 )
− (𝑘2 (𝑅𝑦𝑗 + 1) − 1) ( )−𝛼2 (2.22)
𝑗=1 𝛽
1− 1+
𝑦𝑗

and
( (
)−𝛼3 )
𝛽 𝛽
1+ log 1 +
𝜕 log 𝑙𝑀𝑃𝑆 𝑧1 𝑧1
= log(1 + 𝛽𝑧1−1 ) + ( )−𝛼3
𝜕𝛼3 𝛽
1− 1+
𝑧1
( )−𝛼3 ( ) ( )−𝛼1 ( )
𝛽 𝛽 𝛽 𝛽
𝑚3 1 + log 1 + − 1+ log 1 +
∑ 𝑧𝑙 𝑧𝑙 𝑧𝑙−1 𝑧𝑙−1
+ ( )−𝛼3 ( )−𝛼3
𝛽 𝛽
𝑙=2 1+ − 1+
𝑧𝑙 𝑧𝑙−1

𝑚3
∑ (1 + 𝛽𝑧𝑙−1 ) log(1 + 𝛽𝑧𝑙−1 )
− (𝑘3 (𝑅𝑧𝑙 + 1) − 1) ( )−𝛼3 . (2.23)
𝛽
𝑙=1 1− 1+
𝑧𝑙

The following nonlinear equations cannot be solved analytically to get the estimators 𝛼̂1 , 𝛼̂2 , and 𝛼̂3 of the unknown
parameters 𝛼1 , 𝛼2 , and 𝛼3 . Optimization techniques such as conjugate-gradient or Newton–Raphson optimization
methods are used to compute the estimators of 𝛼1 , 𝛼2 , and 𝛼3 .
 10991638, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/qre.3362 by Central University Of Rajasthan, Wiley Online Library on [04/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAMADAN et al. 9

3 BOOTSTRAP CONFIDENCE INTERVAL

We propose using two additional confidence intervals based on parametric bootstrap methods in this section: (i) percentile
bootstrap method (we name it Boot-p) based on the idea of Efron,47 and (ii) bootstrap-t method (Boot-t) based on the notion
of Hall.48
The two approaches for determining the bootstrap intervals for reliability R are briefly described here.

(i) Boot-p method:

Step 1: Generate random samples (𝑋1;𝑚1 ,𝑛1 ,𝑘1 , 𝑋2;𝑚1 ,𝑛1 ,𝑘1 , … , 𝑋𝑛1 ;𝑚1 ,𝑛1 ,𝑘1 ),
(𝑌1;𝑚2 ,𝑛2 ,𝑘2 , 𝑌2;𝑚2 ,𝑛2 ,𝑘2 , … , 𝑌𝑛2 ;𝑚2 ,𝑛2 ,𝑘2 ), and (𝑍1;𝑚3 ,𝑛3 ,𝑘3 , 𝑍2;𝑚3 ,𝑛3 ,𝑘3 , … , 𝑍𝑛3 ;𝑚3 ,𝑛3 ,𝑘3 ) to calculate 𝛼̂1 , 𝛼̂2 , and 𝛼̂3 .
Step 2: A bootstrap progressive first-failure type-II censored sample,

∗ ∗
(i) (𝑋1;𝑚 , 𝑋2;𝑚 , … , 𝑋𝑛∗ ;𝑚 ,𝑛 ,𝑘 ) is generated from the IL(𝛽, 𝛼1 ) based on the censoring scheme of 𝑅𝑥
1 ,𝑛1 ,𝑘1 1 ,𝑛1 ,𝑘1 1 1 1 1
(ii) (𝑌1;𝑚∗ ,𝑛2 ,𝑘2 , 𝑌2;𝑚 ,𝑛 ,𝑘 , … , 𝑌𝑛∗ ;𝑚 ,𝑛 ,𝑘 ) is generated from the IL(𝛽, 𝛼2 ) based on the censoring scheme of 𝑅𝑦
∗
2 2 2 2 2 2 2 2
∗ ∗
(iii) (𝑍1;𝑚 , 𝑍2;𝑚 , … , 𝑍𝑛∗ ;𝑚 ,𝑛 ,𝑘 ) is generated from the IL(𝛽, 𝛼3 ) based on the censoring scheme of 𝑅𝑧 , then
3 ,𝑛3 ,𝑘3 3 ,𝑛3 ,𝑘3 3 3 3 3
compute the bootstrap sample estimate of 𝑅, say 𝑅̂∗ .

Step 3: Repeat step 2, N number of times.

Step 4: Let 𝐹̂ 1 (𝑥) = 𝑃(𝑅̂ ∗ ≤ 𝑥) denotes the CDF of 𝑅̂ ∗ . Define 𝑅̂ 𝐹(𝐵𝑜𝑜𝑡−𝑝) (𝑥) = 𝐹̂ 1−1 (𝑥).
Then, the approximate 100(1 − 𝑝)% CI for 𝑅̂ is obtained as follows
( ( ) ( ))
𝑃 𝑃
𝑅̂ (𝐵𝑜𝑜𝑡−𝑝) , 𝑅̂ (𝐵𝑜𝑜𝑡−𝑝) 1 − .
2 2

(ii) Boot-t method: The Boot-t method is the same as the Boot-p method.

1. Generate random samples (𝑋1;𝑚1 ,𝑛1 ,𝑘1 , 𝑋2;𝑚1 ,𝑛1 ,𝑘1 , … , 𝑋𝑛1 ;𝑚1 ,𝑛1 ,𝑘1 ), (𝑌1;𝑚2 ,𝑛2 ,𝑘2 , 𝑌2;𝑚2 ,𝑛2 ,𝑘2 , … , 𝑌𝑛2 ;𝑚2 ,𝑛2 ,𝑘2 ) and
(𝑍1;𝑚3 ,𝑛3 ,𝑘3 , 𝑍2;𝑚3 ,𝑛3 ,𝑘3 , … , 𝑍𝑛3 ;𝑚3 ,𝑛3 ,𝑘3 ) to calculate 𝛼̂1 , 𝛼̂2 , and 𝛼̂3 .
∗ ∗
2. Generate a bootstrap sample (𝑋1;𝑚 , 𝑋2;𝑚 , … , 𝑋𝑛∗ ;𝑚 ,𝑛 ,𝑘 ), (𝑌1;𝑚
∗ ∗
, 𝑌2;𝑚 , … , 𝑌𝑛∗ ;𝑚 ,𝑛 ,𝑘 ), and
1 ,𝑛1 ,𝑘1 1 ,𝑛1 ,𝑘1 1 1 1 1 2 ,𝑛2 ,𝑘2 2 ,𝑛2 ,𝑘2 2 2 2 2
∗ ∗
(𝑍1;𝑚 , 𝑍2;𝑚 , … , 𝑍𝑛∗ ;𝑚 ,𝑛 ,𝑘 ).
3 ,𝑛3 ,𝑘3 3 ,𝑛3 ,𝑘3 3 3 3 3
Based on a bootstrap sample estimate 𝑅, say 𝑅̂∗ and the following statistic:
√
𝑚(𝑅̂∗ − 𝑅)
̂
𝑇∗ = √ .
̂ ∗)
𝑉(𝑅

3. Repeat step 2, 𝑁𝑝 number of times.

4. Once 𝑁𝑝 number of 𝑇 ∗ values are obtained, bounds of 100(1 − 𝛾)% confidence interval of 𝑅 are then determined as
√
̂ ∗ )𝐻 −1 (𝑥).
follows: Suppose 𝑇 ∗ follows a cdf given as 𝐻(𝑥) = 𝑃(𝑇 ∗ ≤ 𝑥). For a given 𝑥, define 𝑅(𝐵𝑜𝑜𝑡−𝑡) (𝑥) = 𝑅 + 𝑉(𝑅
The approximate 100(1 − 𝛾)% CI for 𝑅 is then given by the following:
( (𝛾) ( 𝛾 ))
𝑅(𝐵𝑜𝑜𝑡−𝑡) , 𝑅(𝐵𝑜𝑜𝑡−𝑡) 1 − .
2 2

4 BAYESIAN PROCEDURE

This section provides the Bayesian estimate of the unknown parameters 𝛼1 , 𝛼2 , 𝛼3 and Bayesian inference of 𝑅 with respect
to symmetric loss function as squared error (SE) loss functions considering that the three parameters 𝛼1 , 𝛼2 , and 𝛼3 are
random variables. The prior knowledge is incorporated in terms of some prior distributions, and here we assume that the
 10991638, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/qre.3362 by Central University Of Rajasthan, Wiley Online Library on [04/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
10 RAMADAN et al.

three parameters 𝛼1 , 𝛼2 , and 𝛼3 are random variables having independent gamma priors.

𝑎 −1 −𝑏1 𝛼1
𝜋1 (𝛼1 ) ∝ 𝛼1 1 𝑒 , 𝛼1 > 0, 𝑎1 , 𝑏1 > 0,
𝑎 −1 −𝑏2 𝛼2
𝜋2 (𝛼2 ) ∝ 𝛼2 2 𝑒 , 𝛼2 > 0, 𝑎2 , 𝑏2 > 0,
𝑎 −1 −𝑏3 𝛼3
𝜋3 (𝛼3 ) ∝ 𝛼3 3 𝑒 , 𝛼3 > 0, 𝑎3 , 𝑏3 > 0,

where the hyper-parameters 𝑎𝑖 , 𝑏𝑖 , 𝑖 = 1, 2, 3 are assumed to be known and chosen to reflect the previous assumption on
the unknown parameters. As a result, the joint prior density is expressed as

3
∏ 𝑎 −1 −𝑏𝑖 𝛼𝑖
𝜋(𝛼1 , 𝛼2 , 𝛼3 ) = 𝜋1 (𝛼1 ) 𝜋2 (𝛼2 ) 𝜋3 (𝛼3 ) = 𝛼𝑖 𝑖 𝑒 . (4.1)
𝑖=1

The posterior distribution of the parameters 𝛼1 , 𝛼2 , and 𝛼3 denoted by 𝜋∗ (𝛼1 , 𝛼2 , 𝛼3 ∣ 𝑥, 𝑦, 𝑧) up to proportionality can
be obtained by combining the likelihood function Equation (2.4) with the prior via Bayes’ theorem and it can be written
as

𝜋(𝛼1 , 𝛼2 , 𝛼3 ) 𝐿(𝛼1 , 𝛼2 , 𝛼3 ∣ 𝑥, 𝑦, 𝑧)
𝜋∗ (𝛼1 , 𝛼2 , 𝛼3 ∣ 𝑥, 𝑦, 𝑧) = ∞ ∞ ∞ . (4.2)
∫0 ∫0 ∫0 𝜋1 (𝛼1 ) 𝜋2 (𝛼2 ) 𝜋3 (𝛼3 ) 𝐿(𝛼1 , 𝛼2 , 𝛼3 ∣ 𝑥, 𝑦, 𝑧) 𝑑𝛼1 𝑑𝛼2 𝑑𝛼3

The joint posterior to the proportionality can be written as Equation (4.2).

3 𝑚1 ( )−𝛼1 −1
∏ 𝑚𝑗 𝑚𝑗 +𝑎𝑗 −1 𝑚 +𝑚 +𝑚 −𝑏 𝛼 −𝑏 𝛼 −𝑏 𝛼 ∏ 𝛽
𝜋∗ (𝛼1 , 𝛼2 , 𝛼3 ∣ 𝑥, 𝑦, 𝑧) ∝ 𝑘𝑗 𝛼𝑗 𝛽 1 2 3𝑒 1 1 2 2 3 3 𝑥𝑖−2 1 +
𝑗=1 𝑖=1
𝑥𝑖

𝑚2 ( )−𝛼2 −1 ( )𝑘2 (𝑅𝑦𝑗 +1)−1

( ) ∏ 𝛽
−1 −𝛼1 𝑘1 (𝑅𝑥𝑖 +1)−1 −2
1 − (1 + 𝛽𝑥𝑖 ) 𝑦𝑗 1 + 1 − (1 + 𝛽𝑦𝑗−1 )−𝛼2
𝑗=1
𝑦𝑗

𝑚3 ( )−𝛼3 −1
∏ 𝛽 ( )𝑘3 (𝑅𝑧𝑙 +1)−1
𝑧𝑙−2 1+ 1 − (1 + 𝛽𝑧𝑙−1 )−𝛼3 . (4.3)
𝑧𝑙
𝑙=1

Then, the Bayesian estimator of 𝑅 under squared error loss (SEL) function is obtained from the following expression:
∞ ∞ ∞
𝑅̃ = 𝑅 𝜋∗ (𝛼1 , 𝛼2 , 𝛼3 ∣ 𝑥, 𝑦, 𝑧)𝑑𝛼1 𝑑𝛼2 𝑑𝛼3 . (4.4)
∫0 ∫0 ∫0

The integral provided in Equation (4.4) is obviously impossible to calculate analytically. As a result, Metropolis–Hasting
(MH) sampling method is used to construct the Bayesian estimator of 𝑅.

4.1 Metropolis–Hasting sampling method

To derive the Bayesian estimate of 𝑅 and the related credible interval, we now utilize the MH sampling method, which
is a subclass of Monte-Carlo Markov Chain (MCMC) methods. The concept behind this method is to generate posterior
samples of parameters of interest using their posterior conditional density functions. Equation (4.3) gives the posterior
density function of the parameters of interest. From this equation, the posterior conditional density functions of 𝛼1 , 𝛼2 ,
and 𝛼3 can be written as follows:

𝑚1
( −1
)−𝛼1 −1 𝑚1
𝑚 +𝑎 −1
∏ 𝛽 ∏ ( )𝑘1 (𝑅𝑥𝑖 +1)−1
𝜋1∗ (𝛼1 ∣ 𝛼2 , 𝛼3 ) = 𝛼1 1 1 𝑒−𝑏1 𝛼1 1+ 1 − (1 + 𝛽𝑥𝑖−1 )−𝛼1 , (4.5)
𝑖=1
𝑥𝑖 𝑖=1
 10991638, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/qre.3362 by Central University Of Rajasthan, Wiley Online Library on [04/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAMADAN et al. 11

( )−𝛼2 −1 𝑚1 (
𝑚2
∏ 𝛽
−1 ∏ )𝑘2 (𝑅𝑦𝑗 +1)−1
𝑚 +𝑎 −1
𝜋2∗ (𝛼2 ∣ 𝛼1 , 𝛼3 ) = 𝛼2 2 2 𝑒−𝑏2 𝛼2 1+ 1 − (1 + 𝛽𝑦𝑗−1 )−𝛼2 (4.6)
𝑗=1
𝑦𝑗 𝑗=1

and

𝑚3
( −1
)−𝛼3 −1 𝑚1
𝑚 +𝑎 −1 ∏ 𝛽 ∏ ( )𝑘1 (𝑅𝑧𝑙 +1)−1
𝜋3∗ (𝛼3 ∣ 𝛼1 , 𝛼2 ) = 𝛼3 3 3 𝑒−𝑏3 𝛼3 1+ 1 − (1 + 𝛽𝑧𝑙−1 )−𝛼3 . (4.7)
𝑧𝑙
𝑙=1 𝑙=1

It is obvious from Equations (4.5)– (4.7) that the conditional density function of 𝛼1 , 𝛼2 , and 𝛼3 cannot be obtained in the
form of the well-known density functions. In this situation, we can use MH algorithm, proposed by Metropolis et al.49 ,
with normal proposal distribution to generate random samples from the posterior density of 𝛼1 , 𝛼2 , and 𝛼3 , respectively.
The steps of Gibbs sampling are described as follows:

(0) (0) (0)

Step 1. Start with an initial guess (𝛼1 , 𝛼2 , 𝛼3 ) = (𝛼̂1 , 𝛼̂2 , 𝛼̂3 ).
Step 2. Set 𝑡 = 1.
(𝑡) (𝑡) (𝑡) (𝑡−1) (𝑡−1) (𝑡−1)
Step 3. Generate (𝛼1 , 𝛼2 , 𝛼3 ) from 𝜋1∗ (𝛼1 ∣ 𝑥), 𝜋2∗ (𝛼2 ∣ 𝑦) and 𝜋3∗ (𝛼3 ∣ 𝑧) using MH algorithm with the
√ √ √
(𝑡−1) ˆ1 )), 𝑁(𝛼 (𝑡−1) , 𝑣𝑎𝑟(𝛼 ˆ2 )), and 𝑁(𝛼 (𝑡−1) , 𝑣𝑎𝑟(𝛼ˆ3 )), respectively.
proposal distributions 𝑁(𝛼 , 𝑣𝑎𝑟(𝛼 1 2 3

(i) Generate proposals 𝛼1∗ , 𝛼2∗ , and 𝛼3∗ .

(𝑡) (𝑡)
𝜋1∗ (𝛼1 ∣𝑥) 𝜋2∗ (𝛼2 ∣𝑦)
(ii) Evaluate the acceptance probabilities 𝜂𝛼1 = min(1, (𝑡−1) ), 𝜂𝛼2 = min(1, (𝑡−1) ), and 𝜂𝛼3 =
𝜋1∗ (𝛼1 ∣𝑥) 𝜋2∗ (𝛼2 ∣𝑦)
(𝑡)
𝜋3∗ (𝛼3 ∣𝑧)
min(1, (𝑡−1) ).
𝜋3∗ (𝛼3 ∣𝑧)
(iii) Generate 𝑢1 , 𝑢2 , and 𝑢3 from Uniform (0, 1).
(𝑡) (∗) (𝑡) (𝑡−1)
(iv) If 𝑢1 < 𝜂𝛼1 , accept the proposal and set (𝛼1 ) = (𝛼1 ) else, set (𝛼1 ) = (𝛼1 ).
(𝑡) (∗) (𝑡) (𝑡−1)
(v) If 𝑢2 < 𝜂𝛼2 , accept the proposal and set (𝛼2 ) = (𝛼2 ) else, set (𝛼2 ) = (𝛼2 ).
(𝑡) (∗) (𝑡) (𝑡−1)
(vi) If 𝑢3 < 𝜂𝛼3 , accept the proposal and set (𝛼3 ) = (𝛼3 ) else, set (𝛼3 ) = (𝛼3 ).

(𝑡) (𝑡) (𝑡)

Step 4. Compute 𝑅(𝑡) at (𝛼1 , 𝛼2 , 𝛼3 ).
Step 5. Set 𝑡 = 𝑡 + 1.
Step 6. Repeat Steps (3)–(5) 𝑁 times and obtain the posterior sample 𝑅(𝑡) , 𝑡 = 1, … , 𝑁.
∑
The Bayesian estimate of 𝑅 under SEL function, say 𝑅, ̂ is calculated from following expression 𝑅̂ = 1 𝑁 (𝑡)
𝑁 𝑡=𝑀+1 𝑅 .
To construct the credible interval for 𝑅, order 𝑅(𝑡) as 𝑅(1) < 𝑅(2) < … … < 𝑅(𝑁) . Then 100(1 − 𝛾)% credible interval of 𝑅
becomes (𝑅̂ [[𝑁( 𝛾 )]],𝐵 , 𝑅̂ [[𝑁(1− 𝛾 )]],𝐵 ).
2 2
where [[.]] represents the integer value function.

5 SIMULATION

In this section, a Monte Carlo simulation is used to compare the performance of the various methods discussed in the
previous sections. We compare the MLE, MPS, and Bayesian estimation approaches using censoring schemes. Under the
multi-stress–strength model with corresponding choice hyper-parameters in Bayes estimation, we apply the gamma prior
for parameters of IL distribution-based progressive first-failure censoring schemes. For different sample sizes 𝑛 of 15 and
30, and the number of observed failure times 𝑚 of 10, 13, and 15 when n = 15, and 𝑚 =15, 20, 25, and 𝑛 = 30, respectively.
We evaluated two progressive first-failure censoring schemes for showing behavior in various contexts, namely:
Scheme I: 𝑅𝑖 = 0, for 𝑖 ≠ 𝑚 and 𝑅𝑚 = 𝑛 − 𝑚;
Scheme II: 𝑅1 = 𝑛 − 𝑚 and 𝑅𝑖 = 0 for 𝑖 ≠ 1.
 10991638, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/qre.3362 by Central University Of Rajasthan, Wiley Online Library on [04/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
12 RAMADAN et al.

TA B L E 1 MLE, MPS, and Bayesian estimation based on scheme II when 𝛼1 = 0.3, 𝛼2 = 3, 𝛼3 = 5, 𝛽 = 1.5.
MLE MPS Bayesian CI MLE MPS Bayesian
𝒌 𝒏 𝒎 Bias MSE Bias MSE Bias MSE L.ACI L.BP L.BT L.ACI L.CCI
2 15 10 𝛼1 0.0139 0.0055 0.0497 0.0089 0.0396 0.0044 0.2866 0.0180 0.0176 0.3148 0.1876
𝛼2 0.1193 0.4969 −0.0946 0.4349 0.0143 0.0194 2.7249 0.1200 0.1149 2.5595 0.5414
𝛼3 0.1263 0.8299 −0.1869 0.7998 −0.0021 0.0202 3.5384 0.2755 0.2517 3.4301 0.5537
𝑅 −0.0063 0.0027 −0.0189 0.0029 −0.0097 0.0003 0.2017 0.0110 0.0109 0.1991 0.0562
13 𝛼1 0.0121 0.0043 0.0363 0.0089 0.0403 0.0039 0.2539 0.0162 0.0151 0.2902 0.1730
𝛼2 0.0876 0.3989 −0.1021 0.3637 0.0043 0.0117 2.4531 0.0864 0.0979 2.3311 0.4172
𝛼3 0.1029 0.8079 −0.1638 0.7876 −0.0013 0.0118 3.5019 0.1015 0.1010 3.4207 0.4272
𝑅 −0.0058 0.0024 −0.0210 0.0027 −0.0094 0.0002 0.1896 0.0101 0.0092 0.1853 0.0462
15 𝛼1 0.0093 0.0039 0.0347 0.0084 0.0314 0.0026 0.2426 0.0129 0.0119 0.2826 0.1498
𝛼2 0.0799 0.3248 −0.0977 0.2974 0.0082 0.0049 2.2130 0.0796 0.0666 2.1043 0.2758
𝛼3 0.1055 0.7915 −0.1427 0.7514 −0.0021 0.0051 3.4645 0.0811 0.0801 3.3533 0.2711
𝑅 −0.0044 0.0022 −0.0209 0.0025 −0.0076 0.0002 0.1840 0.0092 0.0089 0.1797 0.0383
30 15 𝛼1 0.0053 0.0027 0.0044 0.0027 0.0343 0.0026 0.2037 0.0154 0.0131 0.2313 0.1428
𝛼2 0.0299 0.2504 −0.1247 0.2532 0.0062 0.0190 1.9589 0.1030 0.0912 1.9120 0.5379
𝛼3 0.0268 0.4317 −0.1865 0.4942 0.0039 0.0201 2.5747 0.2073 0.1804 2.6583 0.5550
𝑅 −0.0034 0.0016 −0.0150 0.0018 −0.0080 0.0002 0.1544 0.0096 0.0091 0.1550 0.0484
20 𝛼1 0.0059 0.0025 0.0582 0.0026 0.0372 0.0025 0.1955 0.0066 0.0064 0.2227 0.1360
𝛼2 0.0327 0.2086 −0.1014 0.2096 0.0118 0.0107 1.7868 0.0613 0.0601 1.7508 0.4005
𝛼3 0.0540 0.4014 −0.1333 0.4088 −0.0025 0.0122 2.4759 0.0854 0.0833 2.4524 0.4244
𝑅 −0.0023 0.0014 −0.0166 0.0016 −0.0091 0.0002 0.1443 0.0060 0.0057 0.1419 0.0396
25 𝛼1 0.0060 0.0022 0.0061 0.0025 0.0322 0.0021 0.1807 0.0058 0.0057 0.2132 0.1216
𝛼2 0.0636 0.2019 −0.0572 0.1930 0.0112 0.0052 1.7447 0.0563 0.0571 1.7085 0.2740
𝛼3 0.0342 0.3261 −0.1318 0.3730 −0.0015 0.0052 2.2356 0.0724 0.0706 2.3388 0.2835
𝑅 −0.0044 0.0012 −0.0202 0.0016 −0.0079 0.0001 0.1345 0.0044 0.0044 0.1327 0.0335
4 15 10 𝛼1 0.0059 0.0032 0.0533 0.0069 0.0397 0.0031 0.2217 0.0067 0.0068 0.2491 0.1499
𝛼2 0.0667 0.2949 −0.0900 0.2779 0.0238 0.0195 2.1138 0.0658 0.0659 2.0371 0.5476
𝛼3 0.0577 0.5277 −0.1532 0.5237 −0.0058 0.0203 2.8401 0.0858 0.0819 2.7739 0.5468
𝑅 −0.0041 0.0018 −0.0171 0.0020 −0.0104 0.0003 0.1665 0.0053 0.0054 0.1632 0.0507
13 𝛼1 0.0033 0.0022 0.0488 0.0051 0.0329 0.0021 0.1814 0.0059 0.0059 0.2037 0.1314
𝛼2 0.0381 0.1999 −0.0912 0.1994 0.0161 0.0107 1.7472 0.0560 0.0563 1.7142 0.3924
𝛼3 0.0514 0.3469 −0.1234 0.3756 −0.0025 0.0122 2.3010 0.0751 0.0758 2.3543 0.4363
𝑅 −0.0018 0.0013 −0.0142 0.0014 −0.0084 0.0002 0.1386 0.0043 0.0043 0.1376 0.0407
15 𝛼1 0.0064 0.0020 0.0576 0.0049 0.0299 0.0019 0.1738 0.0054 0.0054 0.1986 0.1189
𝛼2 0.0524 0.1928 −0.0666 0.1889 0.0082 0.0049 1.7099 0.0537 0.0541 1.6846 0.2587
𝛼3 0.0470 0.3416 −0.1254 0.3722 −0.0025 0.0056 2.2849 0.0707 0.0710 2.3417 0.2923
𝑅 −0.0035 0.0012 −0.0175 0.0014 −0.0073 0.0001 0.1364 0.0043 0.0042 0.1345 0.0327
30 15 𝛼1 0.0035 0.0017 0.0395 0.0038 0.0294 0.0013 0.1615 0.0071 0.0073 0.1850 0.1172
𝛼2 0.0546 0.1885 −0.0607 0.1788 0.0145 0.0116 1.6899 0.0733 0.0743 1.6421 0.4091
𝛼3 0.0201 0.3386 −0.1569 0.3403 −0.0014 0.0107 2.2819 0.1023 0.1043 2.2045 0.4162
𝑅 −0.0041 0.0011 −0.0148 0.0013 −0.0075 0.0001 0.1313 0.0059 0.0056 0.1265 0.0374
20 𝛼1 0.0020 0.0014 0.0348 0.0034 0.0261 0.0013 0.1463 0.0065 0.0066 0.1688 0.1075
𝛼2 0.0160 0.1323 −0.0807 0.1330 0.0039 0.0045 1.4260 0.0666 0.0668 1.3955 0.2465
𝛼3 0.0447 0.3054 −0.1028 0.3082 −0.0059 0.0051 2.1614 0.0933 0.0954 2.1408 0.2826
𝑅 −0.0006 0.0010 −0.0132 0.0011 −0.0064 0.0001 0.1213 0.0053 0.0053 0.1183 0.0290
25 𝛼1 0.0029 0.0011 0.0083 0.0031 0.0205 0.0009 0.1300 0.0060 0.0061 0.1507 0.0795
𝛼2 0.0010 0.1193 −0.0829 0.1219 0.0035 0.0012 1.3554 0.0609 0.0603 1.3306 0.1335
𝛼3 0.0383 0.2389 −0.0842 0.2490 0.0005 0.0014 1.9121 0.0836 0.0837 1.9298 0.1504
𝑅 −0.0001 0.0008 −0.0138 0.0009 −0.0048 0.0001 0.1090 0.0051 0.0050 0.1066 0.0206
 10991638, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/qre.3362 by Central University Of Rajasthan, Wiley Online Library on [04/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAMADAN et al. 13

TA B L E 2 MLE, MPS, and Bayesian estimation based on scheme I when 𝛼1 = 0.3, 𝛼2 = 3, 𝛼3 = 5, 𝛽 = 1.5.
MLE MPS Bayesian CI MLE MPS Bayesian
k n m Bias MSE Bias MSE Bias MSE L.ACI L.BP L.BT L.ACI L.CCI
2 15 10 𝛼1 0.0029 0.0035 −0.0349 0.0038 −0.0041 0.0013 0.2310 0.0108 0.0107 0.1996 0.1314
𝛼2 0.0544 0.3211 −0.1199 0.2931 0.0033 0.0108 2.2134 0.0936 0.0915 2.0717 0.3996
𝛼3 0.1239 0.6483 −0.1223 0.5597 0.0038 0.0113 3.1218 0.1366 0.1371 2.8962 0.4221
𝑅 −0.0005 0.0022 0.0066 0.0021 0.0010 0.0001 0.1770 0.0080 0.0081 0.1781 0.0409
13 𝛼1 0.0028 0.0034 0.0228 0.0036 0.0040 0.0012 0.2250 0.0102 0.0101 0.1929 0.1407
𝛼2 0.0312 0.3131 −0.1142 0.2903 −0.0032 0.0100 2.2053 0.0910 0.0903 2.0613 0.2617
𝛼3 0.0917 0.6088 −0.1626 0.5497 −0.0035 0.0105 3.2344 0.1342 0.1334 2.8383 0.2790
𝑅 −0.0026 0.0021 −0.0086 0.0020 −0.0031 0.0001 0.1758 0.0080 0.0080 0.1701 0.0367
30 15 𝛼1 0.0026 0.0019 −0.0217 0.0035 −0.0210 0.0010 0.1685 0.0075 0.0078 0.1271 0.0861
𝛼2 0.0737 0.1893 −0.0353 0.1737 −0.0117 0.0097 1.6826 0.0738 0.0727 1.6293 0.3951
𝛼3 0.0621 0.3331 −0.0795 0.3659 0.0022 0.0091 2.2517 0.1021 0.1060 2.3529 0.4224
𝑅 −0.0035 0.0012 0.0131 0.0015 0.0056 0.0001 0.1327 0.0060 0.0060 0.1435 0.0343
20 𝛼1 0.0046 0.0019 −0.0294 0.0023 −0.0073 0.0007 0.1687 0.0075 0.0076 0.1479 0.0949
𝛼2 0.0424 0.1581 −0.0669 0.1511 −0.0065 0.0050 1.5515 0.0702 0.0703 1.5027 0.2651
𝛼3 −0.0122 0.3307 −0.1767 0.3356 −0.0036 0.0053 2.2782 0.0996 0.0996 2.1649 0.2782
𝑅 −0.0052 0.0011 0.0010 0.0011 0.0019 0.0001 0.1287 0.0059 0.0059 0.1276 0.0295
25 𝛼1 0.0064 0.0018 0.0167 0.0023 0.0069 0.0007 0.1734 0.0074 0.0073 0.1475 0.0938
𝛼2 0.0329 0.1476 −0.0397 0.1492 0.0004 0.0013 1.5666 0.0717 0.0728 1.5075 0.1408
𝛼3 0.0117 0.3059 −0.0909 0.3148 0.0002 0.0013 2.3347 0.0910 0.0909 2.1287 0.1389
𝑅 −0.0036 0.0010 −0.0012 0.0010 −0.0016 0.0000 0.1334 0.0058 0.0059 0.1300 0.0234
4 15 10 𝛼1 0.0032 0.0021 −0.0293 0.0025 −0.0122 0.0018 0.1815 0.0073 0.0082 0.1601 0.1069
𝛼2 0.0537 0.1902 −0.0672 0.1810 −0.0037 0.0114 1.6983 0.0741 0.0805 1.6486 0.4188
𝛼3 0.0518 0.3752 −0.1064 0.3773 0.0051 0.0119 2.3948 0.1059 0.1071 2.3740 0.4069
𝑅 −0.0025 0.0012 0.0041 0.0013 0.0009 0.0001 0.1364 0.0060 0.0060 0.1386 0.0378
13 𝛼1 0.0056 0.0018 0.0176 0.0022 0.0144 0.0011 0.1673 0.0072 0.0078 0.1601 0.1058
𝛼2 0.0483 0.1821 −0.0364 0.1807 0.0014 0.0048 1.6968 0.0738 0.0771 1.6073 0.2652
𝛼3 0.0752 0.3466 −0.0770 0.3691 −0.0100 0.0049 2.2910 0.0992 0.0989 2.3490 0.2648
𝑅 −0.0034 0.0012 −0.0074 0.0013 −0.0038 0.0001 0.1333 0.0060 0.0062 0.1367 0.0316
30 15 𝛼1 0.0034 0.0011 −0.0159 0.0021 −0.0207 0.0007 0.1283 0.0056 0.0057 0.1036 0.0604
𝛼2 0.0125 0.1071 −0.0617 0.1066 −0.0019 0.0042 1.2830 0.0578 0.0561 1.2583 0.2421
𝛼3 0.0252 0.2614 −0.0843 0.2622 0.0104 0.0047 2.0036 0.0918 0.0899 1.9817 0.2598
𝑅 −0.0015 0.0008 0.0125 0.0010 0.0064 0.0001 0.1118 0.0050 0.0050 0.1144 0.0316
20 𝛼1 0.0033 0.0010 −0.0256 0.0016 −0.0074 0.0005 0.1233 0.0055 0.0056 0.1002 0.0637
𝛼2 0.0198 0.0930 −0.0519 0.0898 −0.0041 0.0041 1.1944 0.0549 0.0550 1.1581 0.2407
𝛼3 0.0152 0.2135 −0.1103 0.2105 −0.0029 0.0058 1.8130 0.0819 0.0843 1.7475 0.2938
𝑅 −0.0027 0.0007 0.0029 0.0007 0.0018 0.0000 0.1019 0.0046 0.0046 0.0999 0.0241
25 𝛼1 0.0031 0.0009 0.0119 0.0013 0.0038 0.0004 0.1209 0.0051 0.0050 0.1032 0.0608
𝛼2 0.0158 0.0925 −0.0176 0.0873 0.0026 0.0013 1.1866 0.0545 0.0547 1.1573 0.1416
𝛼3 0.0105 0.2120 −0.0609 0.2061 −0.0007 0.0014 1.8281 0.0804 0.0825 1.7653 0.1444
𝑅 −0.0024 0.0006 −0.0055 0.0006 −0.0020 0.0000 0.0957 0.0042 0.0043 0.0936 0.0200

Several real values of IL distribution-based progressive first-failure censoring approaches are determined using a multi-
stress–strength parameter’s were obtained. Then, we observed three cases for each parameter to investigate the impact of
parameters variations on reliability and model quality. These cases are as follows:

Case 1: 𝛼1 = 0.3, 𝛼2 = 3, 𝛼3 = 5, 𝛽 = 1.5 with 𝑅 = 0.5476.

Case 2: 𝛼1 = 0.5, 𝛼2 = 5, 𝛼3 = 15, 𝛽 = 2 with 𝑅 = 0.6652.
 10991638, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/qre.3362 by Central University Of Rajasthan, Wiley Online Library on [04/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
14 RAMADAN et al.

TA B L E 3 MLE, MPS, and Bayesian estimation based on scheme I when 𝛼1 = 0.5, 𝛼2 = 5, 𝛼3 = 15, 𝛽 = 2.
MLE MPS Bayesian CI MLE MPS Bayesian
k n m Bias MSE Bias MSE Bias MSE L.ACI L.BP L.BT L.ACI L.CCI
2 15 10 𝛼1 0.0211 0.0116 0.0648 0.0165 −0.0006 0.0040 0.4137 0.0189 0.0191 0.4351 0.2392
𝛼2 0.1910 1.0851 −0.1105 0.9401 0.0004 0.0120 3.8721 0.1749 0.1810 3.7455 0.4255
𝛼3 0.0931 3.9861 −0.4268 3.4855 −0.0005 0.0122 7.8256 0.3561 0.3546 7.1317 0.4220
𝑅 −0.0089 0.0012 −0.0014 0.0010 0.0002 0.0001 0.1286 0.0056 0.0058 0.1253 0.0376
13 𝛼1 0.0191 0.0105 −0.0315 0.0089 0.0110 0.0031 0.3948 0.0185 0.0185 0.3494 0.2074
𝛼2 0.2134 1.0193 −0.0801 0.9175 0.0010 0.0050 3.8721 0.1749 0.1810 3.7455 0.2772
𝛼3 0.2216 3.2385 −0.2329 3.1869 −0.0047 0.0054 7.0077 0.3094 0.3222 6.9451 0.2746
𝑅 −0.0067 0.0011 −0.0013 0.0010 −0.0017 0.0001 0.1265 0.0055 0.0054 0.1239 0.0326
30 15 𝛼1 0.0147 0.0056 −0.0954 0.0125 −0.0220 0.0024 0.2871 0.0129 0.0131 0.2301 0.1628
𝛼2 0.0937 0.5742 −0.0936 0.5097 −0.0029 0.0050 2.9507 0.1270 0.1257 2.7773 0.4378
𝛼3 0.0292 2.2855 −0.3311 1.7987 −0.0012 0.0111 5.9310 0.2641 0.2559 5.0996 0.4209
𝑅 −0.0055 0.0007 0.0117 0.0008 0.0035 0.0001 0.1027 0.0045 0.0046 0.0988 0.0282
20 𝛼1 0.0122 0.0053 −0.0277 0.0050 −0.0027 0.0020 0.2807 0.0122 0.0122 0.2549 0.1727
𝛼2 0.0693 0.5034 −0.1128 0.4576 −0.0024 0.0045 2.7707 0.1220 0.1239 2.6171 0.2542
𝛼3 −0.0087 1.9488 −0.3169 1.7669 −0.0061 0.0055 5.4776 0.2471 0.2510 5.0655 0.2768
𝑅 −0.0049 0.0007 0.0013 0.0006 0.0005 0.0000 0.1007 0.0045 0.0045 0.0985 0.0278
25 𝛼1 0.0125 0.0047 0.0086 0.0046 0.0035 0.0016 0.2635 0.0117 0.0118 0.2946 0.1525
𝛼2 0.0403 0.4713 −0.1394 0.4522 −0.0036 0.0035 2.6892 0.1169 0.1151 2.5813 0.2242
𝛼3 0.0829 1.8106 −0.2444 1.6138 −0.0027 0.0039 5.0685 0.2456 0.2495 4.6571 0.2459
𝑅 −0.0034 0.0006 −0.0032 0.0007 −0.0004 0.0000 0.0957 0.0042 0.0043 0.0999 0.0248
4 15 10 𝛼1 0.0084 0.0067 −0.0329 0.0065 0.0006 0.0032 0.3191 0.0141 0.0143 0.2878 0.2147
𝛼2 0.0827 0.6214 −0.1215 0.5803 0.0026 0.0108 2.9125 0.1252 0.1243 2.7677 0.3954
𝛼3 0.0933 2.5408 −0.2894 1.9790 0.0032 0.0124 6.2440 0.2727 0.2703 5.4021 0.4376
𝑅 −0.0038 0.0007 0.0025 0.0006 0.0000 0.0001 0.1061 0.0048 0.0051 0.0991 0.0359
13 𝛼1 0.0078 0.0060 0.0305 0.0062 0.0173 0.0029 0.3034 0.0140 0.0139 0.2732 0.1935
𝛼2 0.0794 0.5572 −0.1239 0.5129 −0.0002 0.0049 3.0761 0.1235 0.1231 2.9509 0.2727
𝛼3 0.1460 2.4466 −0.2520 1.8297 −0.0054 0.0056 6.1108 0.2711 0.2646 5.2864 0.2859
𝑅 −0.0030 0.0007 −0.0030 0.0006 −0.0026 0.0001 0.1065 0.0046 0.0048 0.1025 0.0307
30 15 𝛼1 0.0055 0.0027 −0.0302 0.0060 −0.0262 0.0017 0.2036 0.0094 0.0094 0.1703 0.1234
𝛼2 0.0236 0.2240 −0.0929 0.2215 −0.0084 0.0045 1.7991 0.0806 0.0804 1.7163 0.4061
𝛼3 0.0420 0.8602 −0.0902 0.4284 0.0024 0.0051 3.6356 0.1581 0.1557 2.5439 0.4273
𝑅 −0.0012 0.0003 0.0149 0.0005 0.0043 0.0000 0.0664 0.0030 0.0032 0.0599 0.0224
20 𝛼1 0.0059 0.0028 −0.0294 0.0032 −0.0064 0.0013 0.2066 0.0091 0.0091 0.1882 0.1287
𝛼2 0.0077 0.2174 −0.1116 0.2108 −0.0058 0.0044 1.7893 0.0799 0.0797 1.7048 0.2714
𝛼3 −0.0126 0.7574 −0.1215 0.4319 −0.0065 0.0052 3.4145 0.1564 0.1517 2.5345 0.2829
𝑅 −0.0018 0.0003 0.0050 0.0002 0.0010 0.0000 0.0643 0.0028 0.0028 0.0573 0.0207
25 𝛼1 0.0021 0.0028 0.0283 0.0031 0.0095 0.0009 0.2058 0.0090 0.0089 0.2148 0.1075
𝛼2 0.0073 0.2103 −0.1102 0.2039 0.0011 0.0013 1.8549 0.0833 0.0876 1.8104 0.1294
𝛼3 0.0661 0.5900 −0.0725 0.4228 0.0026 0.0014 3.0029 0.1292 0.1276 2.5356 0.1496
𝑅 −0.0006 0.0002 −0.0043 0.0002 −0.0014 0.0000 0.0586 0.0026 0.0026 0.0550 0.0169

Case 3: 𝛼1 = 0.25, 𝛼2 = 2, 𝛼3 = 15, 𝛽 = 1 with 𝑅 = 0.7729.

MLE estimators of IL distribution based on progressive first-failure censoring approaches are performed by apply-
ing Equations (2.6)–(2.8) in the multi-stress–strength model. Use the Newton–Raphson approach to achieve the best
MLE solution.
 10991638, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/qre.3362 by Central University Of Rajasthan, Wiley Online Library on [04/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAMADAN et al. 15

TA B L E 4 MLE, MPS, and Bayesian estimation based on scheme II when 𝛼1 = 0.5, 𝛼2 = 5, 𝛼3 = 15, 𝛽 = 2.
MLE MPS Bayesian CI MLE MPS Bayesian
k n m Bias MSE Bias MSE Bias MSE L.ACI L.BP L.BT L.ACI L.CCI
2 15 10 𝛼1 0.0155 0.0137 0.0978 0.0270 0.0288 0.0052 0.4559 0.0208 0.0205 0.5184 0.2502
𝛼2 0.1546 1.3301 −0.2025 1.1963 0.0000 0.0104 4.4847 0.1992 0.1972 4.2177 0.3964
𝛼3 0.0149 3.7158 −0.5908 3.9930 −0.0044 0.0129 7.5637 0.3381 0.3335 7.4904 0.4307
𝑅 −0.0083 0.0013 −0.0222 0.0018 −0.0043 0.0001 0.1403 0.0062 0.0064 0.1423 0.0392
13 𝛼1 0.0123 0.0106 0.1211 0.0293 0.0251 0.0037 0.4011 0.0179 0.0177 0.4756 0.2052
𝛼2 0.1131 0.9203 −0.1999 0.8774 0.0022 0.0051 3.7380 0.1663 0.1623 3.5909 0.2762
𝛼3 0.2602 3.7289 −0.3154 3.7270 −0.0087 0.0055 7.5081 0.3362 0.3475 7.4735 0.2793
𝑅 −0.0039 0.0011 −0.0221 0.0016 −0.0039 0.0001 0.1277 0.0055 0.0057 0.1298 0.0325
30 15 𝛼1 0.0140 0.0067 0.0929 0.0172 0.0411 0.0056 0.3153 0.0144 0.0145 0.3642 0.2450
𝛼2 0.0199 0.6846 −0.2358 0.6778 0.0102 0.0115 3.2457 0.1460 0.1460 3.0952 0.4238
𝛼3 0.1252 2.7193 −0.3312 2.8381 −0.0013 0.0136 6.4519 0.2949 0.3038 6.4816 0.4557
𝑅 −0.0033 0.0008 −0.0164 0.0011 −0.0064 0.0001 0.1126 0.0052 0.0052 0.1109 0.0390
20 𝛼1 −0.0002 0.0061 0.0941 0.0174 0.0273 0.0030 0.3076 0.0134 0.0138 0.3618 0.1884
𝛼2 0.1306 0.6109 −0.0938 0.5585 0.0036 0.0043 3.0239 0.1365 0.1407 2.9093 0.2408
𝛼3 0.1881 2.4021 −0.2531 2.5829 −0.0117 0.0055 6.0366 0.2713 0.2744 6.2276 0.2705
𝑅 −0.0024 0.0007 −0.0179 0.0010 −0.0043 0.0001 0.1060 0.0047 0.0047 0.1053 0.0290
25 𝛼1 0.0055 0.0043 0.0869 0.0132 0.0465 0.0050 0.2572 0.0118 0.0118 0.2949 0.1946
𝛼2 0.0414 0.5083 −0.1459 0.4991 0.0142 0.0120 2.7928 0.1240 0.1244 2.7125 0.4296
𝛼3 0.0327 1.8776 −0.3300 2.0151 −0.0022 0.0118 5.3752 0.2397 0.2290 5.4175 0.4214
𝑅 −0.0029 0.0006 −0.0164 0.0009 −0.0073 0.0001 0.0958 0.0042 0.0042 0.0960 0.0301
4 15 10 𝛼1 0.0083 0.0035 0.1056 0.0160 0.0383 0.0033 0.2294 0.0104 0.0103 0.2717 0.1648
𝛼2 0.0025 0.3530 −0.1607 0.3628 0.0088 0.0052 2.3313 0.1060 0.1113 2.2777 0.2835
𝛼3 −0.0051 1.9634 −0.3421 2.0839 −0.0096 0.0053 5.4982 0.2312 0.2388 5.5031 0.2779
𝑅 −0.0027 0.0005 −0.0187 0.0008 −0.0061 0.0001 0.0850 0.0038 0.0039 0.0855 0.0246
13 𝛼1 0.0056 0.0035 0.1012 0.0158 0.0208 0.0012 0.2310 0.0103 0.0103 0.2742 0.1065
𝛼2 0.0319 0.2886 −0.1020 0.2965 0.0035 0.0013 2.1045 0.0962 0.0951 2.0988 0.1389
𝛼3 0.0563 0.9689 −0.2058 1.2651 0.0004 0.0012 3.8562 0.1699 0.1666 4.3391 0.1298
𝑅 −0.0016 0.0003 −0.0187 0.0007 −0.0032 0.0003 0.0713 0.0033 0.0031 0.0770 0.0173
30 15 𝛼1 0.0095 0.0034 0.0869 0.0132 0.0465 0.0050 0.2572 0.0118 0.0100 0.2949 0.1946
𝛼2 0.0414 0.3651 −0.1459 0.4991 0.0142 0.0120 2.7928 0.1240 0.1244 2.7125 0.4296
𝛼3 0.0327 0.8776 −0.1330 1.0151 −0.0022 0.0059 3.3752 0.1540 0.1523 3.3418 0.4214
𝑅 −0.0029 0.0006 −0.0164 0.0009 −0.0073 0.0001 0.0958 0.0042 0.0042 0.0960 0.0301
20 𝛼1 0.0083 0.0033 0.1056 0.0160 0.0383 0.0033 0.2294 0.0104 0.0103 0.2717 0.1648
𝛼2 0.0025 0.3530 −0.1607 0.3628 0.0088 0.0052 2.3313 0.1060 0.1113 2.2777 0.2835
𝛼3 −0.0051 0.7963 −0.3421 0.7084 −0.0096 0.0053 3.3498 0.1423 0.1402 3.3050 0.2779
𝑅 −0.0027 0.0005 −0.0187 0.0008 −0.0061 0.0001 0.0850 0.0038 0.0039 0.0855 0.0246
25 𝛼1 0.0056 0.0032 0.1001 0.0156 0.0208 0.0012 0.2313 0.0103 0.0101 0.2742 0.1065
𝛼2 0.0319 0.2886 −0.1020 0.2965 0.0035 0.0013 2.1045 0.0962 0.0951 2.0988 0.1389
𝛼3 0.0563 0.6895 −0.2058 0.6265 0.0004 0.0012 3.2856 0.1370 0.1365 3.2391 0.1298
𝑅 −0.0016 0.0003 −0.0187 0.0007 −0.0032 0.0000 0.0713 0.0033 0.0031 0.0770 0.0173

Five thousand progressively first-failure-censored samples generated with the IL distribution function, as described by
Balakrishnan and Sandhu.50 For the parameter of the model and 𝑅-stress–strength, we estimate the average bias, MSE of
MLE, MPS, and Bayesian estimator. We used the MCMC technique to produce samples based on 11,000 MCMC simulation
repetitions and remove the first 1000 values as burn-in to avoid any dependency on the beginning value. While the number
of bootstrap samples is 1000 repetitions.
 10991638, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/qre.3362 by Central University Of Rajasthan, Wiley Online Library on [04/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
16 RAMADAN et al.

TA B L E 5 MLE, MPS, and Bayesian estimation based on scheme II when 𝛼1 = 0.25, 𝛼2 = 2, 𝛼3 = 15, 𝛽 = 1.
MLE MPS Bayesian CI MLE MPS Bayesian
k n m Bias MSE Bias MSE Bias MSE L.ACI L.BP L.BT L.ACI L.CCI
2 15 10 𝛼1 0.0076 0.0038 0.0590 0.0086 0.0393 0.0036 0.2389 0.0108 0.0107 0.2808 0.1670
𝛼2 0.0712 0.2039 −0.0738 0.1847 0.0166 0.0102 1.7496 0.0800 0.0829 1.6613 0.3751
𝛼3 0.1164 1.3640 −0.1856 1.5785 −0.0036 0.0118 4.5599 0.2013 0.2013 4.8759 0.4188
𝑅 −0.0054 0.0007 −0.0280 0.0017 −0.0148 0.0005 0.1000 0.0046 0.0047 0.1186 0.0615
13 𝛼1 0.0116 0.0034 0.0578 0.0081 0.0344 0.0025 0.2232 0.0102 0.0102 0.2691 0.1421
𝛼2 0.0759 0.1792 −0.0553 0.1561 0.0085 0.0049 1.6340 0.0751 0.0733 1.5349 0.2654
𝛼3 0.0506 1.2667 −0.1534 1.5772 −0.0044 0.0058 4.4118 0.1966 0.1978 4.8910 0.2818
𝑅 −0.0069 0.0006 −0.0273 0.0016 −0.0130 0.0004 0.0922 0.0041 0.0042 0.1101 0.0528
30 15 𝛼1 0.0091 0.0021 0.0769 0.0082 0.0394 0.0024 0.1749 0.0079 0.0080 0.2099 0.1331
𝛼2 0.0690 0.1310 −0.0417 0.1155 0.0212 0.0048 1.3943 0.0644 0.0665 1.3235 0.3955
𝛼3 0.0213 1.1359 −0.2063 1.4466 −0.0038 0.0057 4.1495 0.1915 0.2000 4.1844 0.4415
𝑅 −0.0057 0.0004 −0.0278 0.0014 −0.0125 0.0004 0.0764 0.0034 0.0035 0.0880 0.0483
20 𝛼1 0.0040 0.0015 0.0618 0.0067 0.0348 0.0021 0.1474 0.0067 0.0067 0.1841 0.1083
𝛼2 0.0557 0.1039 −0.0390 0.0944 0.0092 0.0047 1.2460 0.0515 0.0506 1.1958 0.2674
𝛼3 0.0375 1.1197 −0.2207 1.1859 −0.0036 0.0051 4.1813 0.1852 0.1913 4.6496 0.2716
𝑅 −0.0031 0.0003 −0.0328 0.0014 −0.0123 0.0003 0.0638 0.0028 0.0028 0.0749 0.0408
25 𝛼1 0.0025 0.0014 0.0677 0.0068 0.0233 0.0011 0.1487 0.0065 0.0067 0.1871 0.0924
𝛼2 0.0389 0.0911 −0.0452 0.0844 0.0065 0.0013 1.1745 0.0513 0.0505 1.1264 0.1388
𝛼3 −0.0053 1.0095 −0.2802 1.2481 0.0021 0.0013 3.9425 0.1775 0.1811 4.2435 0.1387
𝑅 −0.0031 0.0003 −0.0301 0.0013 −0.0088 0.0002 0.0653 0.0028 0.0027 0.0773 0.0352
4 15 10 𝛼1 0.0041 0.0020 0.0511 0.0065 0.0351 0.0024 0.1737 0.0078 0.0077 0.2041 0.1221
𝛼2 0.0267 0.1057 −0.0647 0.0997 0.0212 0.0105 1.2618 0.0565 0.0569 1.2044 0.3775
𝛼3 −0.0379 0.9910 −0.2428 0.9629 0.0005 0.0114 4.7303 0.2130 0.2136 4.2955 0.4060
𝑅 −0.0037 0.0003 −0.0235 0.0010 −0.0134 0.0003 0.0714 0.0031 0.0031 0.0824 0.0459
13 𝛼1 0.0052 0.0015 0.0621 0.0060 0.0341 0.0021 0.1514 0.0066 0.0066 0.1828 0.1125
𝛼2 0.0345 0.1046 −0.0699 0.0991 0.0073 0.0050 1.2714 0.0561 0.0565 1.2126 0.2616
𝛼3 0.0350 0.9456 −0.1094 0.9236 −0.0065 0.0052 3.9035 0.1839 0.1843 4.1776 0.4126
𝑅 −0.0040 0.0003 −0.0278 0.0010 −0.0130 0.0003 0.0690 0.0030 0.0029 0.0760 0.0416
30 15 𝛼1 0.0053 0.0012 0.0538 0.0047 0.0353 0.0020 0.1367 0.0059 0.0061 0.1650 0.1024
𝛼2 0.0296 0.0805 −0.0481 0.0766 0.0243 0.0049 1.1070 0.0489 0.0476 1.0697 0.4098
𝛼3 −0.0350 0.9160 −0.2703 0.9021 −0.0030 0.0051 4.1032 0.1781 0.1812 3.7306 0.2670
𝑅 −0.0041 0.0002 −0.0245 0.0009 −0.0135 0.0003 0.0599 0.0027 0.0027 0.0722 0.0380
20 𝛼1 0.0015 0.0009 0.0577 0.0047 0.0322 0.0016 0.1196 0.0053 0.0050 0.1449 0.0862
𝛼2 0.0111 0.0542 −0.0542 0.0538 0.0143 0.0047 0.9126 0.0397 0.0400 0.8852 0.2684
𝛼3 0.0586 0.8243 −0.0902 0.8941 −0.0069 0.0046 3.5552 0.1588 0.1602 3.7908 0.2569
𝑅 −0.0014 0.0002 −0.0242 0.0008 −0.0123 0.0002 0.0539 0.0024 0.0025 0.0637 0.0325
25 𝛼1 0.0045 0.0009 0.0517 0.0046 0.0267 0.0011 0.1148 0.0051 0.0050 0.1395 0.0757
𝛼2 0.0210 0.0533 −0.0345 0.0526 0.0076 0.0013 0.9025 0.0394 0.0397 0.8895 0.1407
𝛼3 −0.0188 0.7115 −0.1657 0.6349 −0.0002 0.0013 3.3091 0.1445 0.1419 3.0582 0.1362
𝑅 −0.0030 0.0002 −0.0278 0.0008 −0.0102 0.0002 0.0513 0.0022 0.0023 0.0573 0.0289

The simulation study’s results are presented in Tables 1–6. Furthermore, we determined the expected length and
95% of various CIs and Bayesian credible intervals in order to study the behavior of different CIs in terms of dif-
ferent sample sizes and varied parameter values, which are shown in tables. From Tables 1–6, we can conclude the
following:
 10991638, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/qre.3362 by Central University Of Rajasthan, Wiley Online Library on [04/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAMADAN et al. 17

TA B L E 6 MLE, MPS, and Bayesian estimation based on scheme I when 𝛼1 = 0.25, 𝛼2 = 2, 𝛼3 = 15, 𝛽 = 1.
MLE MPS Bayesian CI MLE MPS Bayesian
k n m Bias MSE Bias MSE Bias MSE L.ACI L.BP L.BT L.ACI L.CCI
2 15 10 𝛼1 0.0079 0.0029 0.0344 0.0042 0.0200 0.0016 0.2074 0.0094 0.0096 0.1879 0.1225
𝛼2 0.0801 0.1742 −0.0419 0.1518 0.0032 0.0112 1.6075 0.0689 0.0691 1.5200 0.4097
𝛼3 0.1396 1.5262 −0.2051 1.5901 0.0039 0.0119 4.3324 0.1931 0.1976 4.1352 0.4272
𝑅 −0.0051 0.0005 −0.0178 0.0009 −0.0077 0.0002 0.0861 0.0038 0.0037 0.0853 0.0470
13 𝛼1 0.0034 0.0026 −0.0097 0.0024 0.0063 0.0011 0.1997 0.0091 0.0089 0.2161 0.1253
𝛼2 0.0620 0.1548 −0.0589 0.1380 −0.0219 0.0104 1.5248 0.0673 0.0654 1.4394 0.2774
𝛼3 0.0463 1.2385 −0.1807 1.1432 0.0017 0.0110 4.8442 0.1852 0.1952 4.8082 0.2857
𝑅 −0.0037 0.0005 −0.0003 0.0005 −0.0024 0.0002 0.0846 0.0037 0.0037 0.0792 0.0469
30 15 𝛼1 0.0034 0.0013 −0.0094 0.0023 −0.0137 0.0005 0.1417 0.0065 0.0064 0.1155 0.0651
𝛼2 0.0469 0.0887 −0.0297 0.0811 0.0038 0.0050 1.1539 0.0522 0.0546 1.1111 0.3713
𝛼3 0.0109 0.9791 −0.2157 0.8107 −0.0091 0.0052 3.8825 0.1772 0.1788 3.4303 0.4142
𝑅 −0.0031 0.0003 0.0150 0.0004 0.0053 0.0001 0.0624 0.0029 0.0029 0.0582 0.0258
20 𝛼1 0.0058 0.0012 −0.0073 0.0011 0.0020 0.0005 0.1351 0.0062 0.0059 0.1262 0.0854
𝛼2 0.0369 0.0837 −0.0371 0.0779 −0.0018 0.0050 1.1258 0.0500 0.0526 1.0857 0.2709
𝛼3 −0.0824 0.9583 −0.1761 0.8109 −0.0076 0.0046 3.8277 0.1732 0.1795 3.6326 0.2548
𝑅 −0.0045 0.0002 0.0001 0.0002 −0.0008 0.0001 0.0588 0.0026 0.0026 0.0581 0.0335
25 𝛼1 0.0044 0.0011 0.0073 0.0010 0.0122 0.0004 0.1440 0.0065 0.0065 0.1557 0.0813
𝛼2 0.0547 0.0802 −0.0203 0.0722 0.0046 0.0013 1.0903 0.0471 0.0485 1.0511 0.1370
𝛼3 0.0079 1.2615 −0.2147 1.3154 0.0037 0.0015 4.4071 0.1937 0.1917 4.4209 0.1490
𝑅 −0.0033 0.0003 −0.0155 0.0005 −0.0046 0.0001 0.0616 0.0028 0.0027 0.0671 0.0314
2 15 10 𝛼1 0.0035 0.0015 0.0291 0.0026 0.0182 0.0010 0.1511 0.0066 0.0067 0.1624 0.0976
𝛼2 0.0441 0.0946 −0.0399 0.0862 0.0035 0.0088 1.1941 0.0542 0.0532 1.1412 0.3394
𝛼3 0.0794 1.3117 −0.2372 1.1237 0.0049 0.0110 4.4833 0.1990 0.2024 4.0541 0.3909
𝑅 −0.0033 0.0003 −0.0148 0.0005 −0.0071 0.0001 0.0645 0.0028 0.0029 0.0695 0.0369
13 𝛼1 0.0056 0.0014 −0.0087 0.0025 0.0046 0.0005 0.1430 0.0063 0.0062 0.1343 0.0844
𝛼2 0.0403 0.0919 −0.0433 0.0840 −0.0258 0.0086 1.1787 0.0525 0.0524 1.1241 0.2736
𝛼3 0.0070 1.1978 −0.2015 1.0940 0.0069 0.0114 4.2945 0.1979 0.2044 4.0274 0.4157
𝑅 −0.0035 0.0003 0.0004 0.0003 −0.0018 0.0001 0.0636 0.0028 0.0028 0.0632 0.0325
30 15 𝛼1 0.0021 0.0007 −0.0397 0.0020 −0.0133 0.0003 0.1032 0.0045 0.0045 0.0852 0.0488
𝛼2 0.0192 0.0481 −0.0304 0.0454 0.0088 0.0050 0.8570 0.0392 0.0394 0.7921 0.3592
𝛼3 −0.0203 0.5435 −0.0883 0.2358 −0.0097 0.0052 2.8917 0.1283 0.1264 1.8738 0.3909
𝑅 −0.0020 0.0001 0.0147 0.0003 0.0052 0.0001 0.0459 0.0020 0.0020 0.0400 0.0198
20 𝛼1 0.0014 0.0007 −0.0103 0.0008 −0.0005 0.0003 0.1054 0.0046 0.0047 0.1000 0.0612
𝛼2 0.0209 0.0482 −0.0366 0.0421 −0.0055 0.0045 0.8577 0.0386 0.0391 0.8279 0.2362
𝛼3 0.0249 0.4517 −0.0547 0.2297 −0.0087 0.0054 2.6354 0.1181 0.1181 1.8681 0.2928
𝑅 −0.0013 0.0001 0.0030 0.0001 0.0001 0.0001 0.0447 0.0020 0.0020 0.0432 0.0240
25 𝛼1 0.0035 0.0007 0.0102 0.0007 0.0136 0.0006 0.1050 0.0046 0.0046 0.1117 0.0718
𝛼2 0.0134 0.0455 −0.0275 0.0456 0.0030 0.0012 0.8350 0.0378 0.0378 0.8312 0.1337
𝛼3 −0.0042 0.4525 −0.0880 0.3884 0.0028 0.0013 2.6394 0.1170 0.1168 1.8421 0.1376
𝑅 −0.0024 0.0001 −0.0112 0.0001 −0.0052 0.0001 0.0453 0.0021 0.0022 0.0482 0.0277

1. The parameters of the IL distribution under the multi-stress–strength model have a decreasing bias, MSE, and CI
length as the sample size (𝑛) and number of observed failure times (𝑚) rise, according to progressive first-failure
censoring schemes.
2. The parameters of the IL distribution under the multi-stress–strength model have a decreasing bias, MSE, and CI length
as the items inside each group (𝑘) grows, according to progressive first-failure censoring methods.
 10991638, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/qre.3362 by Central University Of Rajasthan, Wiley Online Library on [04/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
18 RAMADAN et al.

TA B L E 7 MLE with SE and different measures.

Estimates KS W AIC BIC
X 𝛼 0.5934 0.2420 0.2162 0.7382 0.0995 0.7013 141.7901 142.3953
𝛽 354.0945 253.3447
Y 𝛼 1.2636 0.7000 0.2879 0.3147 0.1637 1.0749 96.7061 97.3112
𝛽 17.3262 14.1067
Z 𝛼 150.1778 0.2766 0.3604 0.0801 0.5223 79.8178 80.4230
𝛽 0.0671 0.4084

FIGURE 1 Plots of estimated pdfs of distributions for data set with voltage level 42:4 kV. pdfs, probability density functions.

3. When comparing the Bayesian estimates to the MLE and MPS estimates in terms of MSE, the Bayesian estimates always
exceed the MLE and MPS estimates.
4. Scheme I is the best scheme in the majority of scenarios, according to performance indicators.
5. The length of credible CI (L.CCI) is smaller than length of ACI (L.ACI).
6. The length of bootstrap-t (L.BT) CI is smaller than length of bootstrap-p (L.BP).

6 ANALYSIS OF TRANSFORMER INSULATION DATA

This part examines a real data set in order to explain how the methods outlined in the previous sections can be applied
to it. This data set is also used in the multi-stress–strength model to demonstrate the IL distribution-based progressive
first-failure censoring approaches. In Chapter three of Nelson’s book, Nelson,51 the results of a stress–strength life test of
transformer insulation were reported. The test included three voltage levels: 35:4, 42:4, and 46:7 kV, with a typical voltage of
14:4 kV in between. “0.6, 13.4, 15.2, 19.9, 25.0, 30.2, 32.8, 44.4, 50.2+ , and 56.2” are the data when 42:4 kV is used. “40.1, 59.4,
71.2, 166.5, 204.7, 229.7, 308.3, and 537.9” are the values when 35:4 kV is used. “3.1, 8.3, 8.9, 9.0, 13.6, 14.9, 16.1, 16.9, 21.3, and
48.1” are the data when 46:7 kV is used. We want to estimate parameters of IL distribution-based progressive first-failure
censoring techniques under the multi-stress–strength model. Also, we calculate the reliability of the multi-stress–strength
model 𝑃(𝑋 < 𝑌 < 𝑍).
Table 7 presents values of MLE with standard error (SE) and several measures such as the Kolmogorov–Smirnov (KS)
statistic with its P-value, Akaike information criterion (AIC), and Bayesian information criterion (BIC) for the IL distribu-
tion of the Transformer Insulation data set (BIC). The IL distribution matches each data set according to the KS test and
the related P-value. To show how each sample fits into the IL distribution, Figures 1–3 show how each sample fits into the
 10991638, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/qre.3362 by Central University Of Rajasthan, Wiley Online Library on [04/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAMADAN et al. 19

FIGURE 2 Plots of estimated pdfs of distributions for data set with voltage level 35:4 kV. pdfs, probability density functions.

FIGURE 3 Plots of estimated pdfs of distributions for data set with voltage level 46:7 kV. pdfs, probability density functions.

IL distribution.
Table 8 lists the MLE, MPS, and Bayesian estimators of parameter IL distribution-based progressive first-failure censor-
ing procedures by the censored sample under the multi-stress–strength model, with various items within each group. For
MLE and MPS, we notice that the SE of Bayesian estimates is lower than for other models. Under the multi-stress–strength
model, Bayesian estimation is the best performance estimation of parameters for IL distribution-based progressive first-
failure censoring approaches. The Bayesian estimation method’s 𝑃(𝑋 < 𝑌 < 𝑍) dependability is higher than that of the
MLE and MPS methods, proving the stated result.
Convergence diagnostics by trace plot and histogram density estimate of the parameters with the normal curve are
shown in Figure 4. For a total of 10,000 MCMC iterations. Figure 5 shows that the posterior density of MCMC results for
𝜋1 , 𝜋2 , and 𝜋3 has a symmetric normal distribution, respectively. Figures 6 and 7 show that the posterior density of MCMC
results for reliability function has a symmetric normal distribution, respectively.
10991638, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/qre.3362 by Central University Of Rajasthan, Wiley Online Library on [04/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAMADAN et al.

Correlation diagram for posterior distribution when k = 2.

MCMC when k = 2. MCMC, Monte-Carlo Markov chain.
FIGURE 4

FIGURE 5
20
 10991638, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/qre.3362 by Central University Of Rajasthan, Wiley Online Library on [04/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAMADAN et al. 21

T A B L E 8 MLE, MPS, and Bayesian estimation for parameter of IL distribution-based progressive first-failure censoring techniques
under the multi-stress–strength model.
MLE MPS Bayesian
K Estimates SE Estimates SE Estimates SE
1 𝛼1 10.2610 3.2448 10.1388 3.1378 4.9349 1.5042
𝛼2 31.5606 9.9809 31.2859 9.3869 19.9452 5.2552
𝛼3 132.7542 42.1514 134.1648 40.2786 114.1308 20.7094
𝑅 0.5739 0.5771 0.6582
2 𝛼1 12.5979 3.2529 11.4341 3.0311 10.5710 2.9819
𝛼2 32.4805 8.7238 37.2979 8.0128 31.9741 6.4900
𝛼3 130.9766 40.2289 128.2092 39.2743 123.2057 14.6152
𝑅 0.5360 0.5546 0.5586

FIGURE 6 MCMC of R when k = 1. MCMC, Monte-Carlo Markov chain.

FIGURE 7 MCMC of R when k = 2. MCMC, Monte-Carlo Markov chain.

7 CONCLUSION

Recently, the idea of censoring is increasing in popularity. Several researchers have developed various lifetime distributions
and censoring schemes. The IL distribution is one of them, and it is frequently used in reliability and life testing studies.
The progressive first-failure censoring scheme, which is a generalization of the progressive censoring system, was the sub-
ject of this essay. We studied classical and Bayesian estimation methods for dependability within a multi-stress–strength
model. Parameter and reliability values were calculated using ML and MPS. Additionally created were the asymptotic,
boot-p, and boot-t confidence intervals. Using the MCMC technique, Bayesian credible intervals are produced. We use
Monte-Carlo simulation to evaluate the results of various methods. The simulation findings for point estimation show
that, particularly for large sample sizes, the performance of the classical and Bayesian estimators is nearly identical. The
results of the simulation research showed that the MSEs of Bayes estimates for informative priors were much smaller
 10991638, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/qre.3362 by Central University Of Rajasthan, Wiley Online Library on [04/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
22 RAMADAN et al.

than those of the others. Furthermore, the credible intervals for Bayesian estimators are calculated. In most cases, the
coverage chance of Bayesian credible intervals, asymptotic, and boot-p confidence intervals is close to 95%. Furthermore,
bias and MSEs increase along with the total number of samples. To examine the performance of the methods proposed in
this article, a simulated data set and a real data set are analyzed. Also, the diagnostic tests for MCMC technique by trace
plot and histogram density estimate of the parameters and stress–strength reliability function with the normal curves
are introduced.

D A T A AVA I L A B I L I T Y S T A T E M E N T
The data that support the findings of this study are available in the supplementary material of this article.

ORCID
Dina A. Ramadan https://orcid.org/0000-0003-0712-9416
Ehab M. Almetwally https://orcid.org/0000-0002-3888-1275
Ahlam H. Tolba https://orcid.org/0000-0001-8938-9187

REFERENCES
1. Zhou P, Ang B, Poh K. A mathematical programming approach to constructing composite indicators. Ecol Econ. 2007;62(2):291-297.
2. Weerahandi S, Johnson RA. Testing reliability in a stress-strength model when x and y are normally distributed. Technometrics.
1992;34(1):83-91.
3. Surles J, Padgett W. Inference for reliability and stress-strength for a scaled burr type X distribution. Lifetime Data Anal. 2001;7(2):187-200.
4. Al-Mutairi D, Ghitany M, Kundu D. Inferences on stress-strength reliability from Lindley distributions. Commun Stat-Theory Methods.
2013;42(8):1443-1463.
5. Rao GS, Aslam M, Kundu D. Burr-XII distribution parametric estimation and estimation of reliability of multicomponent stress-strength.
Commun Stat-Theory Methods. 2015;44(23):4953-4961.
6. Singh SK, Singh U, Yaday A, Viswkarma P. On the estimation of stress strength reliability parameter of inverted exponential distribution.
Int J Sci World. 2015;3(1):98-112.
7. Almetwally EM, Almongy HM. Parameter estimation and stress-strength model of power Lomax distribution: classical methods and
Bayesian estimation. J Data Sci. 2020;18(4):718-738.
8. Alshenawy R, Sabry M, Almetwally EM, Almongy HM,. Product spacing of stress–strength under progressive hybrid censored for
exponentiated-Gumbel distribution. Comput, Mater Contin. 2021;66(3):2973-2995.
9. Sabry MA, Almetwally EM, Alamri OA, Yusuf M, Almongy HM, Eldeeb AS,. Inference of fuzzy reliability model for inverse Rayleigh
distribution. AIMS Math. 2021;6(9):9770-9785.
10. Abu El Azm WS, Almetwally EM, Alghamdi AS, Aljohani HM, Muse AH, Abo-Kasem O. Stress-strength reliability for exponentiated
inverted Weibull distribution with application on breaking of jute fiber and carbon fibers. Comput Intell Neurosci. 2021;2021:4227346.
11. Chandra S, Owen DB,. On estimating the reliability of a component subject to several different stresses (strengths). Nav Res Logist Q.
1975;22(1):31-39.
12. Yousef MM, Almetwally EM. Multi stress-strength reliability based on progressive first failure for Kumaraswamy model: Bayesian and
non-Bayesian estimation. Symmetry. 2021;13(11):2120.
13. Singh N. On the estimation of p (x1 < y < x2 ). Commun Stat-Theory Methods 1980;9(15):1551-1561.
14. Dutta K, Sriwastav G. An n-standby system with p (x < y < z). IAPQR Trans. 1986;12:95-97.
15. Ivshin V. On the estimation of the probabilities of a double linear inequality in the case of uniform and two-parameter exponential
distributions. J Math Sci. 1998;88(6):819-827.
16. Bhattacharyya G, Johnson RA. Estimation of reliability in a multicomponent stress-strength model. J Am Statist Assoc. 1974;69(348):966-970.
17. Tolba AH, Almetwally EM, Ramadan DA. Bayesian estimation of a one parameter Akshaya distribution with progressively type II censored
data. J Stat Appl Probab Int J. 2022;11:565-579.
18. Ramadan DA, Almetwally EM, Tolba AH. Statistical inference to the parameter of the Akshaya distribution under competing risks data
with application HIV infection to AIDS. Ann Data Sci. 2022;11:1-27.
19. Kotb M, Raqab M. Estimation of reliability for multi-component stress–strength model based on modified Weibull distribution. Stat Pap.
2021;62(6):2763-2797.
20. Maurya RK, Tripathi YM,. Reliability estimation in a multicomponent stress-strength model for Burr XII distribution under progressive
censoring. Braz J Probab Stat. 2020;34(2):345-369.
21. Mahto AK, Tripathi YM, Kızılaslan F. Estimation of reliability in a multicomponent stress–strength model for a general class of inverted
exponentiated distributions under progressive censoring. J Stat Theory Pract. 2020;14(4):1-35.
22. Wang L, Dey S, Tripathi YM, Wu S-J. Reliability inference for a multicomponent stress–strength model based on Kumaraswamy
distribution. J Comput Appl Math. 2020;376:112823.
23. Jha MK, Dey S, Alotaibi, RM, Tripathi YM. Reliability estimation of a multicomponent stress-strength model for unit Gompertz distribution
under progressive Type II censoring. Qual Reliab Eng Int. 2020;36(3):965-987.
 10991638, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/qre.3362 by Central University Of Rajasthan, Wiley Online Library on [04/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAMADAN et al. 23

24. El-Sagheer RM, Tolba AH, Jawa TM, Sayed-Ahmed N. Inferences for stress-strength reliability model in the presence of partially accelerated
life test to its strength variable. Comput Intell Neurosci. 2022;2022:4710536.
25. Rasekhi M, Saber MM, Yousof HM,. Bayesian and classical inference of reliability in multicomponent stress-strength under the generalized
logistic model. Commun Stat-Theory Methods. 2020;50:1-12.
26. Alotaibi RM, Tripathi YM, Dey S, Rezk HR,. Bayesian and non-bayesian reliability estimation of multicomponent stress–strength model
for unit Weibull distribution. J Taibah Univ Sci. 2020;14(1):1164-1181.
27. Maurya RK, Tripathi YM, Kayal T. Reliability estimation in a multicomponent stress-strength model based on inverse Weibull distribution.
Sankhya B. 2021:1-38.
28. Sarhan AM, El-Gohary AI, Tolba AH,. et al. Statistical analysis of a competing risks model with Weibull sub-distributions. Appl Math.
2017;8(11):1671.
29. Sarhan AM, El-Gohary AI, Mustafa and A, Tolba AH,. Statistical analysis of regression competing risks model with covariates using Weibull
sub-distributions. Int J Reliab Appl. 2019;20(2):73-88.
30. Kohansal A, Shoaee S. Bayesian and classical estimation of reliability in a multicomponent stress-strength model under adaptive hybrid
progressive censored data. Stat Pap. 2021;62(1):309-359.
31. Muse AH, Tolba AH, Fayad E, Abu Ali OA, Nagy M, Yusuf M. Modelling the COVID-19 mortality rate with a new versatile modification
of the log-logistic distribution. Comput Intell Neurosci. 2021;2021:8640794.
32. Jana N, Bera S. Interval estimation of multicomponent stress–strength reliability based on inverse Weibull distribution. Math Comput
Simul. 2022;191:95-119.
33. Ahmad HH, Almetwally EM, Ramadan DA. A comparative inference on reliability estimation for a multi-componentstress-strength model
under power Lomax distribution with applications. AIMS Mathematics. 2022;7(10):18050-18079.
34. Ahmad HH, Ramadan DA, Mansour MMM, Aboshady MS. The Reliability of Stored Water Behind Dams Using the Multi-Component
Stress-Strength System. Symmetry. 2023;15:766.
35. Ramadan DA. Assessing the lifetime performance index of weighted Lomax distribution based on progressive type-II censoring scheme
for bladder cancer. Int J Biomath. 2021;14(4):18.
36. Wu S-J, Kuş C. On estimation based on progressive first-failure-censored sampling. Comput Stat Data Anal. 2009;53(10):3659-3670.
37. Elnagar KA, Ramadan DA, El-Desouky BS. Statistical Inference to the Parameter of the Inverse Power Ishita Distribution under Progressive
Type-II Censored Data with Application to COVID-19 Data. J Math Tokushima Univ. 2022;1:1-10.
38. Khalifa EH, Ramadan DA, El-Desouky BS. Estimation to the Parameters of Truncated Weibull Rayleigh Distribution under Progressive
type-II Censoring Scheme with an application to lifetime data. Alfarama J Basic Appl Sci. 2022;3:315-334.
39. Tolba AH, Abushal TA, Ramadan DA. Statistical inference with joint progressive censoring for two populations using power Rayleigh
lifetime distribution. Sci Rep. 2023;13:3832.
40. Kleiber C, Kotz S. Statistical Size Distributions in Economics and Actuarial Sciences. Vol 470. John Wiley & Sons; 2003.
41. Tolba AH, Ramadan DA, Almetwally EM, Jawa TM, Sayed-Ahmed N. Statistical inference for stress-strength reliability using inverse Lomax
lifetime distribution with mechanical engineering applications. Thermal Sci. 2022;26(1):303-326.
42. Ziegel ER. Statistical size distributions in economics and actuarial sciences. Technometrics. 2004;46(4):499.
43. McKenzie D, Miller C, Falk DA. The Landscape Ecology of Fire. Springer Science & Business Media; 2011.
44. Balakrishnan N, Balakrishnan N, Aggarwala R. Progressive Censoring: Theory, Methods, and Applications. Springer Science & Business
Media; 2000.
45. Lawless JF. Statistical Models and Methods for Lifetime Data. Vol 362. John Wiley & Sons; 2011.
46. Washington S, Karlaftis M, Mannering F, Anastasopoulos P. Statistical and Econometric Methods for Transportation Data Analysis.
Chapman and Hall/CRC; 2020.
47. Efron B. The Jackknife, the Bootstrap and Other Resampling Plans. SIAM; 1982.
48. Hall P. Theoretical comparison of bootstrap confidence intervals. Ann Stat. 1988;16:927-953.
49. Metropolis N, Rosenbluth AW, Rosenbluth MN, Teller AH. Equations of state calculations by fast computing machines. J Chem Phys.
1953;21:1087-1092.
50. Balakrishnan N, Sandhu RA. A simple simulational algorithm for generating progressive Type-II censored samples. Am Stat. 1995;49:229-
230.
51. Nelson WB. Accelerated Testing: Statistical Models, Test Plans, and Data Analysis. John Wiley & Sons; 2009.

How to cite this article: Ramadan DA, Almetwally EM, Tolba AH. Statistical inference for multi stress–strength
reliability based on progressive first failure with lifetime inverse Lomax distribution and analysis of transformer
insulation data. Qual Reliab Eng Int. 2023;1-24. https://doi.org/10.1002/qre.3362
 10991638, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/qre.3362 by Central University Of Rajasthan, Wiley Online Library on [04/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
24 RAMADAN et al.

AU T H O R B I O G R A P H I E S

Dina A. Ramadan is an Assistant Professor of Mathematical Statistics at Mathematics Department Faculty of Science
Mansoura University Egypt. She received PhD from the Faculty of Science Mansoura University Egypt in 2016. Her
areas of research where she has several publications in the international journals and conferences include: Statistical
inference, Theory of estimation, Bayesian inference, Theory of reliability, Censored data, Life testing, Competing risk,
Masked data, and Distribution theory.

Ehab M. Almetwally earned the Bachelor degree of Statistics from Faculty of Commerce; Zagazig University, Egypt,
in 2016. Then persuade the post graduate studies at Faculty of Graduate Studies for Statistical Research at Cairo Uni-
versity to earn master of statistics in 2019 with title “Parameter Estimation of Bivariate Models under Some Censoring
Schemes.” Then enrolled at the same faculty to have the PhD degree at statistics to crown the long scientific efforts
by earning the PhD degree in 2023 with title “Contributions to Some Bivariate Models Based on Censored Samples.”
Now, he is 28 years old, and working as an Assistant Professor at Delta University for Science and Technology.

Ahlam H. Tolba is a Lecturer of reliability analysis in the Department of Mathematics, Mansoura University, Egypt.
Her research interests include Statistical Inferences, Reliability Analysis, Generalized Distributions, Shock Models,
Competing Risks Models, and Regression Competing Risks Models. She is a young scholar who published 22 papers
in national and international journals.