Next Article in Journal
Chenciner Bifurcation Presenting a Further Degree of Degeneration
Next Article in Special Issue
An Alternative Approach for Identifying Nonlinear Dynamics of the Cascade Logistic-Cubic System
Previous Article in Journal
Andness Directedness for t-Norms and t-Conorms
Previous Article in Special Issue
Bayesian Influence Analysis of the Skew-Normal Spatial Autoregression Models
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Classical and Bayesian Inference of a Progressive-Stress Model for the Nadarajah–Haghighi Distribution with Type II Progressive Censoring and Different Loss Functions

1
Department of Mathematical Sciences, College of Science, Princess Nourah bint Abdulrahman University, P.O. Box 84428, Riyadh 11671, Saudi Arabia
2
Department of Mathematical Statistical, Faculty of Graduate Studies for Statistical Research, Cairo University, Giza 12613, Egypt
3
Department of Statistical, Faculty of Business Administration, Delta University for Science and Technology, Gamasa 11152, Egypt
4
Department of Management Science and Statistics, The University of Texas at San Antonio, San Antonio, TX 78249, USA
5
Department of Statistics, Al-Azhar University, Cairo 11751, Egypt
*
Author to whom correspondence should be addressed.
Mathematics 2022, 10(9), 1602; https://doi.org/10.3390/math10091602
Submission received: 6 April 2022 / Revised: 26 April 2022 / Accepted: 28 April 2022 / Published: 8 May 2022
(This article belongs to the Special Issue Recent Advances in Computational Statistics)

Abstract

:
Accelerated life testing (ALT) is a time-saving technology used in a variety of fields to obtain failure time data for test units in a fraction of the time required to test them under normal operating conditions. This study investigated progressive-stress ALT with progressive type II filtering with the lifetime of test units following a Nadarajah–Haghighi (NH) distribution. It is assumed that the scale parameter of the distribution obeys the inverse power law. The maximum likelihood estimates and estimated confidence intervals for the model parameters were obtained first. The Metropolis–Hastings (MH) algorithm was then used to build Bayes estimators for various squared error loss functions. We also computed the highest posterior density (HPD) credible ranges for the model parameters. Monte Carlo simulations were used to compare the outcomes of the various estimation methods proposed. Finally, one data set was analyzed for validation purposes.

1. Introduction

Nadarajah and Haghighi [1] proposed a generalization of exponential distribution as an alternative to the regularly used Weibull and gamma distributions in lifetime models. The mode of the Nadarajah and Haghighi (NH) distribution is zero. Recent studies have shown that the NH distribution is extremely adaptable and can be used to model survival data, reliability issues, fatigue life studies, and hydrological data. It can have hazard rate functions that are constant, decreasing, increasing, upside-down bathtub (unimodal), bathtub-shaped, and decreasing–increasing–decreasing; see, for example, [2,3,4,5] and the references given therein.
The NH distribution is said to apply to a random variable T if its probability density function (PDF) is provided by
f ( t ; θ , λ ) = θ λ ( 1 + λ t ) θ 1 e ( 1 ( 1 + λ t ) θ ) ,   θ , λ , t > 0 ,
and the related cumulative density function (CDF) is given by
F ( t ; θ , λ ) = 1 e ( 1 ( 1 + λ t ) θ ) ,   θ , λ , t > 0 ,
where θ and λ are the shape and scale parameters, respectively.
Some individuals or observations may be lost and/or eliminated from the experiment prior to failure in practical applications for reliability and survival analyses. This results in censored data, with no access to the whole data information on failure times for some observations. Progressive type II censoring, first proposed by [6], is one of the most common censoring systems in lifetime scenarios and has received significant attention in the literature. The authors of [7] showed, for example, that in many instances, type II progressive-censoring schemes, which include ordinary type II one-stage right censoring as a particular case, could greatly outperform standard type II censoring. Reference [8] provides a brief overview of the progressive type II censoring technique. Furthermore, we suggest interested readers see [9] for a full examination of the many facets of progressive censoring, ranging from methodology to applications. Mimicking the seminal work of [10], the progressive type II censoring strategy may be represented as follows: Before conducting a life experiment, the experimenter fixes a sample size n, the number of complete observations to be observed, m and progressive-censoring-scheme R 1 ,   R 2 ,     ·   ·   ·   , R m , with n = m + i = 1 m 1 R i . The n units are placed on the life test at the same time. Soon after the initial failure, R 1 surviving units are picked at random and removed from the experiment. Following that, R 2 items are eliminated immediately after the second failure, and so on. This procedure is repeated until all R ι remaining units are deleted following the m-th failure. Note that the R i values are fixed before the experiment begins. Furthermore, if R 1 = R 2 = ·   ·   · = R m = 0 , then n = m which corresponds to the entire sample. Furthermore, if R 1 = R 2 = ·   ·   · = R m 1 = 0 , then R m = n m , which corresponds to the traditional type II right-censoring technique. See [9] for more information on progressive censoring and important references. The authors of [11] propose a new censoring scheme, the type II progressive hybrid censoring scheme, which combines progressive type II and hybrid censoring schemes. A hybrid censoring system is a combination of type I and type II censoring schemes, as defined by [12]. We suggest readers see the book by [9] for a thorough account and results of progressive censoring.
However, in many industrial applications, one of the major requirements for the majority of the production process is that the system is required to operate for a longer period of time. In addition, it is very important to maintain the desired level of system reliability regarding the lifetime of products. In such scenarios, however, life testing under normal stress can lead to a lengthy procedure with expensive cost. Therefore, standard procedures related to progressive censoring may not be advisable. To remedy this situation, the study of accelerated life testing (ALT) has been developed (see [13] for details and the references cited therein). The test procedure makes it possible to quickly obtain information on the life distribution of products by inducing early failure with a much stronger stress than normal. An important added feature in ALT is the step-stress accelerated life test (SSALT). There are two main types of SSALTs: the simple SSALT and the multiple-step SSALT. In the simple SSALT there is a single change of stress during the test. For example, the authors of [14] showed optimum simple SSALT plans in an exponential cumulative exposure (CE) model, the author of [15] studied an exponential CE model with a threshold parameter in the simple SSALT, and the authors of [16] demonstrated optimum modified simple SSALT plans in an exponential CE model, with the consideration that it is desirable to increase the stress at some finite rate. The authors of [17] dealt with a Weibull step-stress model in the simple SSALT. The ALT for modified Kies exponential lifetime distribution based on a progressive-censoring scheme was introduced by [18].
Of particular note is that even though there are some works on complete and censored NH data from both Bayesian and frequentist perspectives, little attention has been paid to the progressive-stress model for the Nadarajah–Haghighi distribution with type II progressive censoring. Accordingly, in this study, we considered classical and Bayesian inference of the NH progressive step-stress model with the SSALT. In addition, we carried out simulation studies to compare the finite sample performance of the estimation methods based on different loss functions.
The rest of the paper is organized as follows. In Section 2, we propose the cumulative exposure model of NH distribution. We discuss the maximum likelihood estimation shape and scale parameters of the NH step-stress model with the SSALT in Section 3. In Section 4, we discuss Bayesian estimation using BSEL and BLINEX loss functions via Markov chain Monte Carlo (MCMC) algorithms. The interval estimation such as asymptotic confidence interval, HPD interval of credibility and Bootstrap confidence intervals is outlined in Section 5. Section 6 describes the simulation studies. An illustrative example is presented in Section 7. Some final remarks are presented in Section 8.

2. A Cumulative Exposure Model of NH Distribution

The basic assumptions under progressive-stress accelerated life testing of the NH distribution are given as follows:
-
Under usual conditions, the lifetime of a unit follows NH( θ ,   λ ).
-
The progressive-stress φ (t) is directly proportional to the time t with constant rate β , i.e., φ ( t ) = β t ,   β > 0 .
-
The scale parameter λ of the CDF in (2) satisfies the inverse power law, as follows
λ ( t ) = 1 a [ φ ( t ) ] b
-
It is assumed that a and b are unknown physical positive parameters and need to be estimated.
-
Assume n is the total number of units tested, φ 0 < φ 1 ( t ) < < φ k ( t ) are the stress levels in the test, and φ 0 is the use stress. Under each progressive-stress level, identical units φ i ( t ) = β i t , i = 1 , 2 , , k ,   n i are tested, and the progressive type II censoring is performed as follows: When the first failure t i 1 : m i : n i   occurs, R i 1 units are picked at random from the remaining n 1 1 surviving units. When the second failure t i 2 : m i : n i occurs, R i 2 items from the remaining n i 2 R i 1 units are withdrawn at random. When the m i t h failure occurs, t i m i : m i : n i   , the test is terminated, and all remaining R i m i : m i : n i   = n i m i j = 1 m i 1 R i m i items are removed.
-
The complete samples and type II censored samples are clearly specific examples of this technique. Under the progressive-stress φ i ( t ) , the observed progressive-censoring data are t i 1 : m i : n i   < t i 2 : m i : n i   < · · · < t i m i : m i : n i   , i = 1 , 2 , , k .
-
The linear cumulative exposure model (CEM) accounts for the effect of changing stress; for more details, see [13].
-
The CDF in progressive stress, φ i ( t ) , and the linear cumulative exposure model is given as follows
G i ( t ) = F i ( t ) ,   i = 1 , 2 , , k ,
where Δ t = 0 t 1 λ i ( u ) d u = a β i b t b + 1 b + 1 , F (·) denotes the CDF of the NH distribution under progressive-stress φ i ( t ) with scale parameter λ. Thus,
G i ( t ) = 1 e ( 1 ( 1 + a β i b t b + 1 b + 1 ) θ ) ,   t > 0 ,   a ,   b ,   θ > 0 ,   i = 1 , 2 , , k .
The corresponding PDF is given by
g i ( t ) = a θ β i b t b ( 1 + a β i b t b + 1 b + 1 ) θ 1 e [ 1 ( 1 + a β i b t b + 1 b + 1 ) θ ] ,
where t > 0 ,   a ,   b ,   θ > 0 ,

3. Maximum Likelihood Estimation

In this section, we discuss the maximum likelihood estimates (MLEs) for the parameters a, b, and θ under the progressive-stress ALT based on progressively type II censored data. We assume that the observed data under the stress levels φ i ( t ) , i = 1 , , k and j = 1 , , m i ,   . The likelihood function is obtained as
L ( a , b , θ ) = i = 1 k c i j = 1 m i ,   g i ( t i j ) [ 1 G i ( t i j ) ] R i j ,   t i j > 0 ,
where c i = n i ( n i 1 R i 1 ) ( n i 2 R i 2 ) ( n i m i + 1 j = 1 m i 1 R i j ) . By substituting (3) and (4) in (5), we obtain
L ( a , b , θ ) = i = 1 k c i j = 1 m i ,   a θ β i b t i j b ( 1 + a β i b t i j b + 1 b + 1 ) θ 1 e [ 1 ( 1 + a β i b t i j b + 1 b + 1 ) θ ] { e [ 1 ( 1 + a β i b t i j b + 1 b + 1 ) θ ] } R i j .
Then,
L ( a , b , θ ) = i = 1 k c i j = 1 m i ,   a θ β i b t i j b ( 1 + a β i b t i j b + 1 b + 1 ) θ 1 { e [ 1 ( 1 + a β i b t i j b + 1 b + 1 ) θ ] } R i j + 1
The logarithm of (6) is given by
l o g   L ( a , b , θ ) = i = 1 k l o g   c i + i = 1 k   m i ( l o g a + l o g θ ) + b i = 1 k   m i l o g   β i + b i = 1 k j = 1 m i l o g ( t i j ) + ( θ 1 ) i = 1 k j = 1 m i l o g ( 1 + a β i b t i j b + 1 b + 1 ) + i = 1 k j = 1 m i ( R i j + 1 ) [ 1 ( 1 + a β i b t i j b + 1 b + 1 ) θ ] .
By taking the first partial derivatives of (7) with respect to a, b, and θ and then equating each to zero, we obtain
l o g   L   a = i = 1 k m i a + ( θ 1 ) i = 1 k j = 1 m i β i b t i j b + 1 ( b + 1 ) ( 1 + a β i b t i j b + 1 b + 1 ) + i = 1 k j = 1 m i θ β i b t i j b + 1 ( b + 1 ) ( R i j + 1 ) ( 1 + a β i b t i j b + 1 b + 1 ) θ 1 = 0 ,
l o g   L   b = i = 1 k m i l o g β i + i = 1 k j = 1 m i t i j + ( θ 1 ) i = 1 k j = 1 m i a β i b t i j b + 1 [ l o g β i b + 1 + l o g t i j b + 1 1 ] ( 1 + a β i b t i j b + 1 b + 1 ) + i = 1 k j = 1 m i θ ( R i j + 1 ) [ 1 a β i b t i j b + 1 [ l o g β i b + 1 + l o g t i j b + 1 1 ] ( b + 1 ) 2 ] = 0 ,
l o g   L   θ = i = 1 k m i θ + θ i = 1 k j = 1 m i l o g ( 1 + a β i b t i j b + 1 b + 1 ) i = 1 k j = 1 m i θ β i b t i j b + 1 ( b + 1 ) ( R i j + 1 ) ( 1 + a β i b t i j b + 1 b + 1 ) θ l o g ( 1 + a β i b t i j b + 1 b + 1 ) = 0 .
Equating the first partial derivatives of (7) to zero and solving the resulting Equations (8)–(10) numerically, we could numerically obtain the MLEs a ^ , b ^ and θ ^ for the three unknown parameters a, b, and θ. As a result, numerical solutions to the nonlinear system in (8)–(10) can be found using an iterative approach such as Newton–Raphson.

4. Bayesian Estimation

In this section, we study the Bayes estimates of the parameters a, b, and θ using the balanced squared error loss (BSEL) and balanced linear-exponential (BLINEX) error loss function. It is well known that Bayesian analysis begins with the prior specification for the unknown parameters, and in the absence of prior knowledge, noninformative priors are often preferred to minimize the impacts of the prior distributions. Thus, in this study, we assumed that the three parameters are independent, and the noninformative priors of a, b, and θ are given as
π ( Θ ) = π ( a , b , θ ) = π 1 ( a ) π 2 ( b ) π 3 ( θ ) 1 a b θ , a , b , θ > 0 ,
where Θ is the parameter model. The resulting joint posterior density function is obtained by
π * ( a , b , θ | t ) π ( a , b , θ ) L ( a , b , θ ) ,   which   shows   that π * ( a , b , θ | t ) 1 a b θ i = 1 k c i

Bayes Estimation Using BSEL and BLINEX Loss Functions

We followed the work of [19,20,21] and determined that the Bayes parameter estimators of a function w using the BSEL function are given by
w ^ B S E L = ω w ^ M L + ( 1 ω ) Θ w ^ M L   π * ( Θ | t )   d Θ
where w ^ M L is the MLE of w , and 0 < ω < 1 and π * ( . | t ) is defined in (12). Similarly, the Bayes parameter estimator of a function w using the BLINEX loss function is obtained as
w ^ B L I N E X = 1 γ l n [ ω e γ w ^ M L + ( 1 ω ) Θ e γ w ^ M L π * ( Θ | t )     d Θ ]
where γ   0 is the shape parameter of the BLINEX loss function. We used the MCMC algorithm to solve the integrals in (13) and (14). The conditional posterior distributions of the parameters a, b and θ are obtained as follows
π * ( a | b , θ , t ) 1 a i = 1 k j = 1 m i   β i b t i j b ( 1 + a β i b t i j b + 1 b + 1 ) θ 1 { e [ 1 ( 1 + a β i b t i j b + 1 b + 1 ) θ ] } R i j + 1 ,
π * ( b | a , θ , t ) 1 b i = 1 k j = 1 m i   β i b t i j b ( 1 + a β i b t i j b + 1 b + 1 ) θ 1 { e [ 1 ( 1 + a β i b t i j b + 1 b + 1 ) θ ] } R i j + 1 ,
and
π * ( θ | b , a , t ) 1 θ i = 1 k j = 1 m i   β i b t i j b ( 1 + a β i b t i j b + 1 b + 1 ) θ 1 { e [ 1 ( 1 + a β i b t i j b + 1 b + 1 ) θ ] } R i j + 1 .
Because it is difficult to obtain the conditional posterior distributions of the parameters a ,   b , and θ in the form of some well-known distributions, we employed the Metropolis–Hastings (MH) algorithm for generating posterior samples for the above conditional posterior distributions to obtain the appropriate Bayes estimates of the unknown parameters. However, drawing samples directly from the conditional posterior distributions of model parameters may be problematic. As a result, we need to choose appropriate known distributions to obtain these conditional distributions. More information about this topic can be found in [22,23,24,25,26,27]. In this study, we followed the steps of the MH algorithm to create a random sample from these conditional distributions:
Step 1:
For the parameters ( a ,   b , θ ), set the initial guess to ( a 0 ,   b 0 , θ 0 ).
Step 2:
Set j = 1.
Step 3:
Create a ~ N ( a j , σ 11 ) , b ~ N ( b j , σ 22 ) and θ ~ N ( θ j , σ 33 ) , where σ is the variance–covariance matrix.
Step 4:
Compute p = π ( a j ,   b j , θ j | x _ ) π ( a j 1 ,   b j 1 , θ j 1 | x _ ) .
Step 5:
With probability m i n ( 1 , p ) , accept ( a j ,   b j , θ j ),
Step 6:
To obtain B number of samples for the parameters ( a ,   b , θ ), repeat steps (3) to (5) B times.

5. Interval Estimation

We created confidence intervals (CIs) utilizing the MLE’s normality criterion using the asymptotic CI (ACI) as well as the highest posterior density (HPD) intervals for the parameters a ,   b , and θ .

5.1. Asymptotic Confidence Interval

The ACI was obtained using the normality property of MLEs with the parameters a, b , and θ . It is known that the MLEs of ( a ,   b , θ ) have an asymptotic distribution under some regularity conditions, such that
( a   ^ a ) ~ N ( 0 , σ 11 ) ,   ( b ^ b ) ~ N ( 0 ,   σ 22 ) ,   ( θ ^ θ ) ~ N ( 0 ,   σ 33 ) ,  
where the variance–covariance matrix for the parameters ( a ,     b , θ ) is σ i i . The 100 ( 1 ω ) % percent confidence intervals for the parameters ( θ 1 ,   θ 2 , θ 3 ) = ( a ,   b ,   θ ) are written as ( θ ^ i Z 1 ω 2 σ i i ) , i = 1, 2, 3, where σ i i indicates the ( i , i ) t h the variance–covariance matrix σ and ( θ ^ 1 a ^ ,   θ ^ 2 b ^ , θ ^ 3 θ ^ ) .

5.2. HPD Interval of Credibility

The HPD credible interval ( L , U ) was constructed for a random quantity θ * that is derived by debating the following expression.
p ( L θ * U ) = L U π * ( θ * | t ) d θ * = 1 γ
Because finding the interval ( L , U ) analytically is difficult, we used the posterior samples acquired in Section 4, to obtain the requisite HPD credible intervals using the technique described in [28].

5.3. Bootstrap Confidence Intervals

The CIs based on parametric bootstrap sampling utilizing the percentile intervals were produced; for additional information, see [29]. The bootstrap CIs were calculated by using the following algorithm:
(1)
Calculate the MLE values of the parameters using the original data θ , a , and b .
(2)
To make a bootstrap sample t * , use the variables θ ^ , a ^ , and b ^ .
(3)
The bootstrap estimates θ ^ * , a ^ * , and b ^ * , respectively, are obtained based on t * .
(4)
To obtain the bootstrap samples, repeat steps 1–3 several times and organize each estimate in ascending order
{ θ ^ * [ 1 ] ,   θ ^ * [ 2 ] , ,   θ ^ * [ I ] } , { a ^ * [ 1 ] ,   a ^ * [ 2 ] , ,   a ^ * [ I ] } and   { b ^ * [ 1 ] ,   b ^ * [ 2 ] , , b * [ I ] } ,  
The 100 ( 1 δ ) % percentile bootstrap CIs for ω are then calculated as follows:
( ω ^ i L ,   ω ^ i U ) = ( ω ^ i * [ δ 2 I ] ,   ω ^ i * [ ( 1 δ 2 ) I ] ) , i = 1 , 2 , 3 ,
where ω ^ 1 * = θ ^ * ,   ω ^ 2 * = a ^ * and ω ^ 3 * = b ^ * .

6. Simulation Study

We carried out a simulation study according to the following algorithm to obtain the likelihood and Bayesian estimation of the NH parameters based on progressive stress and study their properties through the mean squared error (MSE), bias, and length of confidence intervals.
Step 1:
Using the algorithm presented in [10], k 2 progressively type II censored random samples are generated from the uniform (0,1) distribution ( U i 1 , U i 2 , , U i m i , ) , for given values of m i , ,   i = 1 , 2 , , k .
Step 2:
To compare the performance of the estimation procedures developed in the study, we consider the following two schemes for each stress:
Scheme 1: R i m i = n i m i , R i j = 0 ; j = 1 , , m i 1 .
Scheme 2: R i 1 = n i m i , R i j = 0 ; j = 1 , , m i .
Step 3:
Progressively type II censored random samples ( t i 1 , , t i m i ) are produced, and from inverse CDF (3), we specify the values of parameters as follows:
In Table 1  ( θ = 1.7 ,   a = 1.3 ,   b = 2 ) ,   k = 2 ,   and   β 1 = 40 ,   β 2 = 80 .
In Table 2  ( θ = 0.8 ,   a = 0.5 ,   b = 1.3 ) ,   k = 2 , and β _ 1   = 40 ,   β _ 2   = 80 .
In Table 3  ( θ = 3 ,   a = 2.5 ,   b = 0.6 ) ,   k = 2 , and β 1 = 40 ,   β 2 = 80 .
In Table 4 ( θ = 1.7 ,   a = 1.3 ,   b = 2 ) ,   k = 4 , and β 1 = 40 ,   β 2 = 80 ,   β 3 = 110 ,   β 4 = 150 .
In Table 5 ( θ = 0.8 ,   a = 0.5 ,   b = 1.3 ), k = 4 , and β 1 = 40 ,   β 2 = 80 ,   β 3 = 110 ,   β 4 = 150 .
In Table 6 ( θ = 3 ,   a = 2.5 ,   b = 0.6 ) , k = 4 , and β 1 = 40 ,   β 2 = 80 ,   β 3 = 110 ,   β 4 = 150 .
Step 4:
The MLEs ( θ ^ ,   a ^ ,   b ^ ) are obtained numerically by solving the likelihood equations with respect to ( θ ,   a ,   b ) in (8)–(10) by using an iterative Newton–Raphson algorithm using the maxlik function of the “maxlik” package in the R program; for more information in this topic see [30].
Step 5:
Based on (15)–(17), and the MH algorithm, the Bayesian estimations with the BSEL and BLINEX loss functions of the parameters ( θ ,   a ,   b ) are computed by (13) and (14), respectively.
Step 6:
The above steps are repeated I times based on I different samples, and then the average of likelihood and Bayesian estimations are computed, with their MSE, bias, and length of confidence intervals (LCI) of the parameters ( θ ,   a ,   b ).
Step 7:
In length of CI (LCI) of the MLE of each parameter, we compute the ACI for likelihood estimators and bootstrap CIs with the percentile algorithm and t algorithm, which can be denoted as LBP and LBT, respectively. In the LCI of Bayesian estimation, we compute the HPD for each loss function, denoted by the LCCI.
Table 1. MLE and Bayesian estimation with different loss functions in the simple ramp when θ = 1.7, a = 1.3, b = 2.
Table 1. MLE and Bayesian estimation with different loss functions in the simple ramp when θ = 1.7, a = 1.3, b = 2.
k = 2θ = 1.7, a = 1.3, b = 2MLEBSELBLINEX (c = −0.5)BLINEX (c = 0.5)MLEBSELBLINEX
(c = −0.5)
BLINEX
(c = 0.5)
n1, n2Schemer BiasMSEBiasMSEBiasMSEBiasMSELCILBPLBPLCCILCCILCCI
20, 15165%θ−0.67511.5008−0.19380.0886−0.11720.0801−0.26190.10994.00940.13310.12710.85920.98210.7748
a0.38241.5335−0.05210.07880.02090.0994−0.11740.07404.61930.14320.14531.00481.11450.8922
b0.06730.6760−0.29720.1145−0.26160.0964−0.33140.13453.21390.09980.09930.63150.65510.6181
85%θ−0.48790.7308−0.08790.0657−0.00970.0736−0.15810.07242.75290.08990.08940.95521.05240.8647
a0.50080.99840.05040.06120.01950.08900.04260.06913.39110.11050.11010.98210.98340.9084
b0.06250.6469−0.16220.0558−0.12600.0471−0.19670.06673.10260.09780.09650.67180.69340.6551
265%θ0.08720.56330.05500.05830.13910.0909−0.02060.04502.92350.08940.08750.92621.04700.8281
a−0.27250.51040.05170.10580.12620.1418−0.01580.08662.59020.08050.08181.20141.30561.0823
b0.50350.85940.06070.04110.10110.05000.02180.03603.05290.09080.09150.73050.74900.7048
85%θ0.08230.50780.05420.05800.13460.0879−0.01710.04482.83450.08040.08030.91370.91030.8448
a−0.18010.49020.08630.10410.16110.15090.01850.07782.65340.08580.08601.09771.22371.0492
b0.32570.55930.04940.03630.08790.04370.01260.03232.64040.08540.08450.71490.73840.6916
40, 50165%θ−0.96781.4864−0.31390.1585−0.24890.1345−0.37260.19003.77650.12500.11740.92071.00060.8607
a0.31261.00500.12600.12440.19740.17540.06010.09134.44080.14330.13731.21161.33911.0992
b−0.02620.5537−0.45340.2334−0.42620.2114−0.47910.25582.91670.09380.09290.65670.67740.6325
85%θ−0.95311.0410−0.21430.1193−0.14880.1099−0.27380.13801.42820.04510.04551.04191.14710.9624
a0.30330.94270.12300.12020.13740.17010.05820.09023.16090.10340.10271.20281.19081.1697
b0.24680.3993−0.25970.1017−0.23100.0904−0.28690.11452.28160.07200.07200.73460.75320.7076
265%θ0.01890.51950.03500.06600.10670.0902−0.03070.05572.82590.08750.08780.96621.06010.9024
a−0.10430.34580.05340.10260.11540.1347−0.00340.08342.26990.07490.07501.13261.21501.0415
b0.29880.33160.08060.04160.11040.04890.05190.03651.93090.06100.06100.70830.71810.6939
85%θ0.01990.29020.03450.06270.10230.0901−0.01550.05472.11120.06910.06890.95210.96020.9010
a−0.00410.26440.04990.10130.11060.13250.00460.08292.01680.06060.06171.12281.13441.1043
b0.13560.15060.04370.03320.07110.03810.01730.03021.42620.04750.04760.66480.68110.6482
Table 2. MLE and Bayesian estimation with different loss functions in the simple ramp when θ = 0.8, a = 0.5, b = 1.3.
Table 2. MLE and Bayesian estimation with different loss functions in the simple ramp when θ = 0.8, a = 0.5, b = 1.3.
k = 2θ = 0.8, a = 0.5, b = 1.3MLEBSELBLINEX (c = −0.5)BLINEX (c = 0.5)MLEBSELBLINEX
(c = −0.5)
BLINEX
(c = 0.5)
n1, n2Schemer BiasMSEBiasMSEBiasMSEBiasMSELCILBPLBPLCCILCCILCCI
20, 15165%θ−0.67182.8664−0.15680.1203−0.12740.1278−0.18450.11516.09490.20600.19730.99501.08390.9307
a0.24401.62980.14660.12610.17360.14960.12020.10504.91460.15490.15541.09241.14381.0436
b0.09570.4839−0.07380.0664−0.02680.0708−0.11740.06752.70250.08580.08480.95781.02050.8990
85%θ−0.25150.1789−0.05860.1017−0.02620.1189−0.08890.08931.33360.04160.04141.07671.13270.9832
a0.19990.26320.16470.12490.19030.14680.13980.10541.85320.05980.06071.11271.15711.0571
b0.24940.31020.00440.06170.04800.0721−0.03630.05661.95320.05940.05790.95741.01150.8972
265%θ0.19970.27010.11530.10810.15260.13380.08010.08761.88190.06040.06021.08371.13690.9657
a−0.14900.13010.05860.08550.08160.09930.03650.07371.28840.04090.03991.02311.08040.9543
b0.26800.29610.09330.06650.13420.08230.05510.05571.85740.05930.05990.92710.98370.8748
85%θ0.16730.21620.08450.04950.11210.06210.05860.04001.70140.05220.05160.77960.84720.7347
a−0.06990.09760.06640.05810.08500.06690.04860.05081.19410.03790.03800.82330.86640.7746
b0.15210.21920.03720.02300.06270.02730.01300.02061.73670.05830.05880.57650.60310.5623
40, 50165%θ−0.68093.2570−0.36950.1771−0.35280.1723−0.38490.18226.55500.22540.20220.60220.66130.5605
a0.62601.10680.30230.19560.32780.22380.27690.16923.31630.11540.10661.15251.20971.0901
b0.20430.17520.04350.09650.09450.1144−0.00560.08491.43260.04730.04691.18001.24111.1403
85%θ−0.44150.2093−0.26210.0908−0.25000.0879−0.27330.09420.47080.01530.01510.51550.54530.4948
a0.43690.34650.27010.14010.29030.15850.25020.12311.54700.04740.04790.96611.00580.9208
b0.25430.14670.02830.03400.05540.03900.00220.03111.12290.03490.03490.70750.72350.6965
265%θ0.08720.15980.10370.08660.13370.10450.07530.07201.53000.04680.04720.95171.02250.9087
a−0.10120.06630.02970.05190.04520.05800.01490.04670.92820.02770.02760.83750.86770.7991
b0.19050.21010.07960.06390.11030.07570.05060.05521.63500.05350.05260.89250.94350.8638
85%θ0.08390.12480.07570.04920.09730.05930.05500.04091.34620.04230.04260.76450.81620.7216
a−0.04840.04730.05790.04890.07170.05500.04460.04350.83110.02650.02640.70490.73580.6820
b0.09610.12940.02430.02590.04330.02860.00590.02411.35940.04320.04320.63070.65170.6262
Table 3. MLE and Bayesian estimation with different loss functions in the simple ramp when θ = 3, a = 2.5, b = 0.6.
Table 3. MLE and Bayesian estimation with different loss functions in the simple ramp when θ = 3, a = 2.5, b = 0.6.
k = 2 θ = 3, a = 0.5, b = 0.6MLEBSELBLINEX (c = −0.5)BLINEX (c = 0.5)MLEBSELBLINEX
(c = −0.5)
BLINEX
(c = 0.5)
n1, n2Schemer BiasMSEBiasMSEBiasMSEBiasMSELCILBPLBPLCCILCCILCCI
20, 15165%θ−1.00811.4542−0.34990.1451−0.26930.0983−0.42420.20082.59560.08510.08470.58260.61820.5601
a0.06791.2060−0.37240.1673−0.29380.1205−0.44380.22214.29870.13860.14100.64870.72080.6095
b−0.11760.1318−0.26130.0778−0.25210.0737−0.27020.08191.34730.04370.04330.36030.37300.3459
85%θ−0.44540.4194−0.12020.0252−0.07330.0168−0.16520.03751.84400.05910.05910.39910.40980.3920
a−0.06490.2608−0.11680.0311−0.06700.0235−0.16390.04331.98650.06540.06370.52810.54300.5133
b−0.05810.0912−0.16320.0352−0.15550.0331−0.17070.03741.16200.03670.03740.35680.36240.3499
265%θ−0.15330.4265−0.01280.04100.08080.0511−0.09950.04842.48970.07340.07260.80160.83790.7820
a−0.24000.6128−0.01000.05970.08770.0755−0.09870.06392.92240.09010.08980.95701.02320.9205
b0.22400.18300.02900.02700.04380.02930.01450.02531.42910.04270.04260.61810.63020.6014
85%θ−0.00760.28890.01330.01520.06320.0198−0.03460.01582.10770.06620.06510.47800.48120.4698
a−0.13650.32280.02530.02690.08030.0347−0.02680.02552.16290.06790.06840.63790.66990.6282
b0.12880.11350.00580.01440.01550.0150−0.00390.01401.22060.03860.03840.44620.45340.4430
40, 50165%θ−1.19732.0957−0.43880.2123−0.36330.1551−0.50820.27613.19170.10210.09960.55240.58500.5390
a0.42421.8779−0.42400.2012−0.35660.1528−0.48520.25435.11060.16390.16430.56460.61970.5296
b−0.30400.1460−0.34200.1222−0.33670.1188−0.34710.12550.90770.03100.03050.26630.27150.2631
85%θ−1.11301.6602−0.16450.0369−0.12190.0252−0.20530.05172.54590.07910.07920.39120.39500.3890
a0.33560.3522−0.11700.0290−0.07450.0218−0.15750.03934.91140.15090.15040.47520.48740.4623
b−0.13760.0633−0.24160.0645−0.23720.0625−0.24590.06650.82640.02680.02650.29280.29640.2898
265%θ−0.07360.3191−0.00930.04210.07860.0525−0.08990.04772.19640.06880.06810.81180.83960.7883
a−0.07310.29960.01950.05350.10480.0712−0.05850.05172.12750.06740.06590.88910.95020.8511
b0.10740.05930.03510.01800.04380.01900.02640.01710.85760.02880.02850.47910.48150.4721
85%θ−0.01280.25890.00970.01670.05570.0204−0.03440.01731.30100.05900.05890.50140.51220.4937
a0.06940.27070.02600.03000.07310.0363−0.01880.02841.34290.06120.06120.66000.68070.6407
b0.06470.04090.00310.00990.00900.0101−0.00270.00970.75130.02400.02410.39420.39550.3921
Table 4. MLE and Bayesian estimation with different loss functions in multi ramp when θ = 1.7, a = 1.3, b = 2.
Table 4. MLE and Bayesian estimation with different loss functions in multi ramp when θ = 1.7, a = 1.3, b = 2.
k = 4θ = 1.7, a = 1.3, b = 2MLEBSELBLINEX (c = −0.5)BLINEX (c = 0.5)MLEBSELBLINEX
(c = −0.5)
BLINEX
(c = 0.5)
n1, n2,
n3, n4
Schemer BiasMSEBiasMSEBiasMSEBiasMSELCILBPLBPLCCILCCILCCI
20, 15, 18, 10165%θ−0.87151.0614−0.22420.1068−0.15190.0943−0.28930.13022.15610.09650.09650.89560.96320.8047
a0.80961.48460.00010.08820.06880.1142−0.06250.07653.78220.16910.16941.10781.25370.9950
b0.11570.3767−0.30890.1190−0.28010.1037−0.33650.13552.36510.10700.10660.61090.62910.5869
85%θ−0.55900.7077−0.10890.0307−0.06770.0259−0.14760.03882.46670.11440.11350.52400.55970.4972
a0.63540.90140.04700.04020.09050.05390.00620.03233.06550.14410.13850.67950.72530.6408
b0.09620.3299−0.18720.0476−0.17040.0419−0.20360.05382.22210.10030.10170.44370.44360.4419
265%θ0.14630.48470.03730.07560.11700.1094−0.03450.06172.94520.13150.12950.99351.08400.8752
a−0.31370.37550.02280.11390.08990.1463−0.03880.09592.06580.09460.09371.22441.38131.1134
b0.35910.41380.09130.04370.12250.05190.06130.03792.09440.09770.09730.72760.74070.7164
85%θ0.13310.35520.03060.02260.07380.0306−0.01000.01922.86900.12160.12260.55570.60390.5304
a−0.14760.31430.04600.04160.08820.05330.00670.03482.12220.10060.09920.75280.80650.7120
b0.21910.25860.02380.01390.04190.01540.00600.01321.80080.07610.07640.44270.45140.4406
25, 20, 20, 25165%θ−1.03711.2431−0.27410.1331−0.20600.1158−0.33560.16061.60590.07650.07650.91311.03960.8033
a0.97940.94510.06840.11740.13670.15810.00520.09172.75290.12550.12141.11291.22471.0286
b0.20070.4168−0.33600.1360−0.30960.1204−0.36120.15252.40780.09990.10010.54560.54770.5355
85%θ−0.83270.8583−0.16660.0526−0.12760.0443−0.20310.06381.59330.07190.07240.62020.65240.5959
a0.87460.83240.08590.05080.12820.06620.04640.04062.93300.13220.12750.79320.85150.7411
b0.21370.3082−0.18740.0491−0.17190.0440−0.20270.05472.01060.08730.08730.46300.47160.4563
265%θ0.01050.44560.04410.07540.11720.1023−0.02320.06282.61890.11690.11711.00201.09860.9428
a−0.18690.3433−0.00150.10930.05590.1339−0.05450.09552.17880.09300.09301.16951.26021.0874
b0.30250.26930.09810.04300.12470.05040.07250.03741.65430.07170.07130.69760.72070.6846
85%θ0.02240.21490.02030.02380.06300.0310−0.01960.02121.81680.08210.08220.61060.66410.5580
a−0.07160.21820.05360.04870.09290.06040.01680.04151.81150.08090.08100.76610.81930.7406
b0.13560.13200.02840.01750.04440.01910.01270.01641.32230.05920.06010.48370.49100.4805
Table 5. MLE and Bayesian estimation with different loss functions in multi ramp when θ = 0.8, a = 0.5, b = 1.3.
Table 5. MLE and Bayesian estimation with different loss functions in multi ramp when θ = 0.8, a = 0.5, b = 1.3.
k = 4θ = 0.8, a = 0.5, b = 1.3MLEBSELBLINEX (c = −0.5)BLINEX (c = 0.5)MLEBSELBLINEX
(c = −0.5)
BLINEX
(c = 0.5)
n1, n2,
n3, n4
Schemer BiasMSEBiasMSEBiasMSEBiasMSELCILBPLBPLCCILCCILCCI
20, 15, 18, 10165%θ−0.46260.2492−0.22340.1145−0.19910.1159−0.24610.11510.73630.03440.03280.82270.85690.7749
a0.39130.39150.20190.14630.22700.17030.17720.12401.91580.08330.08391.12131.18351.0524
b0.18800.1968−0.05470.0723−0.01070.0790−0.09590.07031.57650.07390.07370.95011.01670.8885
85%θ−0.29080.1596−0.11930.0548−0.10070.0563−0.13700.05451.07470.05060.04980.71170.74820.6735
a0.23220.21340.16030.09240.17950.10580.14150.08041.56690.06880.06790.84840.88280.8237
b0.16230.1711−0.03660.0257−0.01330.0263−0.05880.02631.49280.06850.06730.57390.58800.5635
265%θ0.12990.17480.13740.12530.17160.15230.10440.10221.55940.07330.07261.08831.17171.0231
a−0.14730.08580.01640.05960.03390.0670−0.00050.05360.99320.04290.04340.88810.92470.8439
b0.22400.21530.08170.06310.11260.07430.05250.05511.59450.06980.06980.91480.95420.8811
85%θ0.11530.15220.08090.05050.10360.06060.05920.04211.47240.07200.07100.70570.75040.6762
a−0.07630.07750.06790.05940.08420.06720.05200.05261.05090.04660.04650.81910.84630.7834
b0.11430.15200.02340.02410.04290.02640.00460.02281.46270.06160.06140.58700.59480.5762
25, 20, 20, 25165%θ−0.51910.2829−0.29370.1456−0.27450.1430−0.31170.14900.45510.02120.01940.80070.86420.7501
a0.51510.48800.22510.14200.24730.16160.20320.12371.85170.08450.08431.04371.07770.9973
b0.16940.1096−0.00740.08470.03340.0947−0.04650.07891.11610.04880.04751.12511.19541.0986
85%θ−0.38310.1711−0.21760.0699−0.20510.0671−0.22950.07290.61140.02850.02770.52990.55130.5081
a0.31780.22740.21120.10110.22990.11510.19280.08831.39470.06430.06370.85760.89270.8125
b0.14900.10180.01340.02870.03670.0318−0.00890.02691.01240.04500.04200.65880.66960.6415
265%θ0.10170.16930.12410.09180.15470.11150.09510.07551.56430.07060.06901.00201.07910.9475
a−0.15490.0652−0.01930.0454−0.00550.0490−0.03260.04260.79690.03540.03530.75450.77830.7351
b0.21070.18700.08830.05660.11500.06550.06290.04991.48210.07000.06980.84340.85940.8319
85%θ0.09920.12700.08020.04290.09920.05000.06190.03691.34280.05950.05950.71240.75470.6790
a−0.05490.05200.04440.04170.05770.04650.03140.03740.86860.03940.03920.68760.72380.6680
b0.08870.10770.02850.02410.04550.02660.01200.02251.23990.05510.05520.59330.60700.5709
Table 6. MLE and Bayesian estimation with different loss functions in multi ramp when θ = 3, a = 2.5, b = 0.6.
Table 6. MLE and Bayesian estimation with different loss functions in multi ramp when θ = 3, a = 2.5, b = 0.6.
k = 4θ = 3, a = 2.5, b = 0.6MLEBSELBLINEX (c = −0.5)BLINEX (c = 0.5)MLEBSELBLINEX
(c = −0.5)
BLINEX
(c = 0.5)
n1, n2,
n3, n4
Schemer BiasMSEBiasMSEBiasMSEBiasMSELCILBPLBPLCCILCCILCCI
20, 15, 18, 10165%θ−1.31502.3042−0.33310.1341−0.25180.0897−0.40780.18762.97400.09310.09340.58580.62350.5653
a0.89893.0227−0.34910.1514−0.27370.1096−0.41770.20095.83650.18310.18190.65460.70890.6228
b−0.09130.0606−0.26220.0764−0.25520.0731−0.26910.07980.89710.02750.02710.32340.33090.3182
85%θ−0.47080.4836−0.11160.0243−0.06540.0166−0.15570.03582.00720.06720.06610.41920.43110.4174
a0.11340.3691−0.09820.0274−0.05100.0217−0.14300.03722.34080.07690.07650.52180.54810.5058
b−0.11240.0513−0.18000.0387−0.17490.0370−0.18520.04050.77090.02470.02470.29340.29410.2908
265%θ−0.10240.4661−0.01980.04130.07170.0510−0.10430.04862.64730.08610.08600.80060.83130.7613
a−0.16800.7442−0.01160.06680.08030.0824−0.09560.06933.31860.10900.10960.99761.06730.9505
b0.14890.07590.03310.01670.04260.01780.02370.01590.90860.02840.02850.48470.49060.4771
85%θ0.09480.40530.01630.01640.06500.0211−0.03030.01662.48200.08080.08060.49260.50310.4903
a−0.13870.39140.03110.02910.08230.0368−0.01740.02712.39250.07650.07720.65870.68360.6349
b0.07540.05010.00020.00830.00610.0085−0.00570.00820.82690.02630.02730.35910.36160.3564
25, 20, 20, 25165%θ−1.09131.4984−0.37550.1641−0.29850.1155−0.44660.22062.17510.06710.06680.58900.61910.5681
a0.17530.6860−0.36360.1586−0.29290.1171−0.42820.20693.17480.10030.10060.62940.66720.6001
b−0.16340.0712−0.28650.0879−0.28090.0849−0.29200.09090.82720.02540.02530.29290.29750.2878
85%θ−0.69340.7946−0.13390.0307−0.08910.0214−0.17680.04352.19720.06890.06800.43720.44880.4279
a0.14420.6040−0.09500.0278−0.04920.0224−0.13850.03723.06040.10130.09110.54720.56770.5313
b−0.11700.0424−0.18940.0414−0.18510.0398−0.19360.04290.66420.01970.01990.27970.27910.2786
265%θ−0.18620.6036−0.03000.04190.05940.0488−0.11230.05102.95820.09560.09610.79940.82090.7782
a−0.04280.6685−0.03070.05990.05490.0696−0.10880.06563.20220.10080.10160.93971.00070.9127
b0.14380.06340.04410.01470.05150.01570.03670.01390.81050.02610.02610.43170.43650.4286
85%θ−0.04430.11860.01140.01770.05870.0219−0.03400.01821.33940.04270.04320.52420.52700.5207
a−0.06030.12820.02880.03090.07790.0382−0.01790.02901.38430.04350.04310.65070.69200.6365
b0.04850.02370.00340.00710.00820.0072−0.00140.00700.57320.01830.01810.31870.32060.3177

Simulation Results

Based on the two censoring schemes, considering two cases of levels of the stress (simple ramp when k = 2 and multi ramp when k = 4   (k > 2)), simulation results are presented in Table 1, Table 2, Table 3, Table 4, Table 5 and Table 6. First, Table 1, Table 2 and Table 3 show the likelihood and Bayesian estimations of the parameters ( θ ,   a ,   b ) in the simple ramp with their bias, MSE, and LCI for different sample sizes ( n i ), and number of failures ( m i ) which can be obtained by rounding r × n i . Second, Table 4, Table 5 and Table 6 show the likelihood and Bayesian estimations of the parameters ( θ ,   a ,   b ) in the multi-ramp with their bias, MSE, and LCI for different sample sizes ( n i ), and number of failures ( m i ). Bayesian estimations of the parameters based on the BLINEX loss function, considering two values of c (c = −0.5, 0.5), are also included. The number of MH iterations of the algorithm is 10,000, and the number of simulations is 1000. The coverage probabilities of CI were set to 95%.
From the simulation results presented in Table 1, Table 2, Table 3, Table 4, Table 5 and Table 6, the following points can be observed:
(1)
For fixed values of the sample sizes n i , by increasing the censored sample sizes, m i , the bias, MSE, and LCI of the estimates decrease for the two different censored schemes.
(2)
For fixed values of m i , by increasing the sample sizes n i , the bias, MSE, and LCI decrease for different censored schemes.
(3)
For fixed values of m i or n i or scheme, by increasing the level of stress k, the bias, MSE, and LCI decrease.
(4)
For fixed values of m i or n i or scheme, we note that Scheme 2 is better than Scheme 1 for some or all parameters.
(5)
The bias and MSE reduce significantly, and the symmetric and asymmetric Bayesian estimations are better than the MLE in the considered scenarios.
(6)
The LCI reduces significantly, the symmetric and asymmetric Bayesian estimations of the HPD are better than the ACI of MLE.
(7)
We observe that the shortest lengths of the CI are the bootstrap CI.

7. An Illustrative Example

We used a practical experiment to test the MH distribution’s superiority over its competitors. In this case, [13] employed a previous dependability experiment. This data set contains stress time from ramp-voltage studies with small light bulbs. In the ramp-voltage trials, 62 light bulbs were utilized in the first level of stress, and 61 light bulbs were used in the second level of stress, with ramp rates of 2.01 and 2.015 V/h, respectively, as well as a 2 V experimental design stress. The data are for a complete sample that has not been censored as follows:
The first level under progressive stress 2.01 V/h, t 1 , and the second level under progressive stress 2.015 V/h, t 2 , are shown in Figure 1 and Figure 2, respectively.
To see if the real data set follows the NH distribution, we utilized the Kolmogorov–Smirnov distance (KSD) test to obtain the p-value for making decisions. Table 7 contains the MLEs for the NH distribution with standard error (SE) and Akaike information criteria (AIC). Figure 1 and Figure 2 show the fitted CDF, estimated PDF, and P-P plots for each stress level of real data. These figures confirmed that the NH distribution is a fit for each stress level of this real data. According to the results of KSD and P-values for the NH distribution in Table 7 and Figure 1 and Figure 2, we conclude that the NH distribution is suitable for fitting each stress level of real data. If the P-value of the KS test is more than 0.05, then we may conclude that the distribution is a fit for this data. When the P-value of the KS test increases, the line of estimated cdf is near the empirical cdf, and the line of estimated pdf has drowned the histogram of data. Table 7 shows the KS test values and the P-value of the KS test < 0.5 as shown in Figure 1 and Figure 2. Table 8 shows the MLEs and Bayesian estimates for the parameters of the progressive-stress model and the MLEs and Kolmogorov–Smirnov test for each stress level of real data. Table 8 results were obtained after those of Table 7.
As illustrated in Figure 3, we see that Bayes estimators outperform MLEs in most cases. The estimated parameters maximize the log-likelihood function. The roots were computed, and they invariably pinpoint the global maximum rather than the local maximum. By displaying the log-likelihood function, we were able to confirm our findings. As seen by the blue dot, the estimate is at its maximum location along the curve.
Figure 4 and Figure 5 illustrate the trace plots and normal curves of the posterior distribution for MCMC estimation, respectively. These numbers also showed that the MCMC sequences converged quickly to become stationary. Furthermore, as shown in Figure 6, we used the random drawings to generate random draws for the sub-survivors and overall survivors as well as the Bayesian estimators and 95% credible intervals for those functions at various times. The normal distribution is often used to estimate the conditional posterior distributions of the parameters because they are generally symmetric and unimodal (see Figure 5). Figure 6 indicate that Bayes point and interval estimates of the sub-survivor functions and the overall survivor function.

8. Concluding Remarks

In this work, we studied classical and Bayesian estimation procedures for the progressive-stress model for the NH distribution with type II progressive-censoring data. Specifically, we first considered the MLEs of the NH parameters and adopted their asymptotic normal distribution to construct approximate confidence intervals. Then, we studied Bayesian estimators and the corresponding HPD intervals of the unknown parameters with suitable prior distributions and different loss functions. Numerical results from simulation studies and real data examples showed that the performance of the proposed methods is satisfactory for statistical inference of the step-stress accelerated model with the NH distribution based on progressive type II censored data.

Author Contributions

Conceptualization, R.A., E.M.A. and H.R.; Funding acquisition, R.A.; Methodology, R.A. and H.R.; Resources, F.S.A.; Software, E.M.A.; Supervision, M.W.; Writing—original draft, R.A. and H.R.; Writing—review & editing, F.S.A., E.M.A. and M.W. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by Princess Nourah bint Abdulrahman University Researchers Supporting Project number (PNURSP2022R50), Princess Nourah bint Abdulrahman University, Riyadh, Saudi Arabia.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

The data used to support the findings of this study are included within the article.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Nadarajah, S.; Haghighi, F. An extension of the exponential distribution. Statistics 2011, 45, 543–558. [Google Scholar] [CrossRef]
  2. Khan, M.N.; Saboor, A.; Cordeiro, G.M.; Nazir, M. A weighted Nadarajah and Haghighi distribution. UPB Sci. Bull. Ser. A Appl. Math. Phys. 2018, 80, 133–140. [Google Scholar]
  3. Tahir, M.H.; Cordeiro, G.M.; Ali, S.; Dey, S.; Manzoor, A. The inverted Nadarajah–Haghighi distribution: Estimation methods and applications. J. Stat. Comput. Simul. 2018, 88, 2775–2798. [Google Scholar] [CrossRef]
  4. Chesneau, C.I.; Okorie, E.I.; Bakouch, H.S. A Skewed Nadarajah–Haghighi Distribution with Some Applications. J. Indian Soc. Probab. Stat. 2020, 21, 225–245. [Google Scholar] [CrossRef]
  5. Mansoor, M.; Tahir, M.H.; Alzaatreh, A.; Cordeiro, G.M. The Poisson Nadarajah—Haghighi Distribution: Properties and Applications to Lifetime Data. Int. J. Reliab. Qual. Saf. Eng. 2020, 27, 2050005. [Google Scholar] [CrossRef]
  6. Cohen, A.C. Progressively Censored Samples in Life Testing. Technometrics 1963, 5, 327. [Google Scholar] [CrossRef]
  7. Hofmann, G.; Cramer, E.; Balakrishnan, N.; Kunert, G. An asymptotic approach to progressive censoring. J. Stat. Plan. Inference 2005, 130, 207–227. [Google Scholar] [CrossRef]
  8. Lawless, J.F. Statistical Models and Methods for Lifetime Data; John Wiley and Sons: Hoboken, NJ, USA, 2003. [Google Scholar]
  9. Balakrishnan, N.; Aggarwala, R. Progressive Censoring: Theory, Methods, and Applications; Springer: Berlin/Heidelberg, Germany, 2000. [Google Scholar]
  10. Balakrishnan, N.; Sandhu, R.A. A simple simulation algorithm for generating progressive Type-II censored samples. Am. Stat. 1995, 199549, 229–230. [Google Scholar]
  11. Kundu, D.; Joarder, A. Analysis of Type-II progressively hybrid censored data. Comput. Stat. Data Anal. 2006, 50, 2509–2528. [Google Scholar] [CrossRef]
  12. Gupta, R.D.; Kundu, D. Hybrid censoring schemes with exponential failure distribution. Commun. Stat.-Theory Methods 1998, 27, 3065–3083. [Google Scholar] [CrossRef]
  13. Nelson, W.B. Accelerated Testing: Statistical Models, Test Plans, and Data Analysis; Wiley: New York, NY, USA, 1990. [Google Scholar]
  14. Miller, R.; Nelson, W. Optimum Simple Step-Stress Plans for Accelerated Life Testing. IEEE Trans. Reliab. 1983, R-32, 59–65. [Google Scholar] [CrossRef]
  15. Xiong, C. Step Stress Model with Threshold Parameter. J. Stat. Comput. Simul. 1999, 63, 349–360. [Google Scholar] [CrossRef]
  16. Park, S.-J.; Yum, B.-J. Optimal design of accelerated life tests under modified stress loading methods. J. Appl. Stat. 1998, 25, 41–62. [Google Scholar] [CrossRef]
  17. Lu, M.W.; Rudy, R.J. Step-stress accelerated test. Int. J. Mater. Prod. Technol. 2002, 17, 425–434. [Google Scholar] [CrossRef]
  18. El-Raheem, A.M.A.; Almetwally, E.M.; Mohamed, M.S.; Hafez, E.H. Accelerated life tests for modified Kies exponential lifetime distribution: Binomial removal, transformers turn insulation application and numerical results. AIMS Math. 2021, 6, 5222–5255. [Google Scholar] [CrossRef]
  19. Ahmadi, J.; Jozani, M.J.; Marchand, E.; Parsian, A. Bayes estimation based on k-record data from a general class of distributions under balanced type loss functions. J. Stat. Plan Inference 2009, 139, 1180–1189. [Google Scholar] [CrossRef]
  20. Ahmadi, J.; Jozani, M.J.; Marchand, E.; Parsian, A. Prediction of k-records from a general class of distributions under balanced type loss functions. Metrika 2008, 70, 19–33. [Google Scholar] [CrossRef]
  21. Ahmed, E.A. Bayesian estimation based on progressive Type-II censoring from two-parameter bathtub-shaped lifetime model: An Markov chain Monte Carlo approach. J. Appl. Stat. 2014, 41, 752–768. [Google Scholar] [CrossRef]
  22. Contreras-Reyes, J.E. Lerch distribution based on maximum nonsymmetric entropy principle: Application to Conway’s game of life cellular automaton. Chaos Solitons Fractals 2021, 151, 111272. [Google Scholar] [CrossRef]
  23. Almongy, H.; Alshenawy, F.; Almetwally, E.; Abdo, D. Applying Transformer Insulation Using Weibull Extended Distribution Based on Progressive Censoring Scheme. Axioms 2021, 10, 100. [Google Scholar] [CrossRef]
  24. Bantan, R.; Hassan, A.S.; Almetwally, E.; Elgarhy, M.; Jamal, F.; Chesneau, C.; Elsehetry, M. Bayesian Analysis in Partially Accelerated Life Tests for Weighted Lomax Distribution. Comput. Mater. Contin. 2021, 68, 2859–2875. [Google Scholar] [CrossRef]
  25. Alotaibi, R.; Al Mutairi, A.; Almetwally, E.M.; Park, C.; Rezk, H. Optimal Design for a Bivariate Step-Stress Accelerated Life Test with Alpha Power Exponential Distribution Based on Type-I Progressive Censored Samples. Symmetry 2022, 14, 830. [Google Scholar] [CrossRef]
  26. Alotaibi, R.; Khalifa, M.; Almetwally, E.M.; Ghosh, I.; Rezk, H. Classical and Bayesian Inference of a Mixture of Bivariate Exponentiated Exponential Model. J. Math. 2021, 2021, 1–20. [Google Scholar] [CrossRef]
  27. Tolba, A.H.; Almetwally, E.M.; Ramadan, D.A. Bayesian Estimation of A one Parameter Akshaya Distribution with Progressively Type II Censord Data. J. Stat. Appl. Probab. 2022, 11, 565–579. [Google Scholar]
  28. Chen, M.H.; Shao, Q.M. Monte Carlo estimation of Bayesian credible and HPD intervals. J. Comput. Graph. Stat. 1999, 8, 69–92. [Google Scholar]
  29. Tibshirani, R.; Efron, B. An Introduction to the Bootstrap; Chapman & Hall, Inc.: New York, NY, USA, 1993. [Google Scholar]
  30. R Core Team. A Language and Environment for Statistical Computing; R Foundation for Statistical Computing: Vienna, Austria, 2021. [Google Scholar]
Figure 1. Fitted CDF with empirical CDF, estimated PDF, and P-P plots for first stress levels.
Figure 1. Fitted CDF with empirical CDF, estimated PDF, and P-P plots for first stress levels.
Mathematics 10 01602 g001
Figure 2. Fitted CDF with empirical CDF, estimated PDF, and P-P plots for second stress levels.
Figure 2. Fitted CDF with empirical CDF, estimated PDF, and P-P plots for second stress levels.
Mathematics 10 01602 g002
Figure 3. Profile–likelihood for the three parameters for the real data set. Where ge-03 is g × 10 3 and g is value of parameter g without virgule.
Figure 3. Profile–likelihood for the three parameters for the real data set. Where ge-03 is g × 10 3 and g is value of parameter g without virgule.
Mathematics 10 01602 g003
Figure 4. Iterations of MCMC results for progressive-stress model. Where ge-06 is g × 10 6 and g is value of parameter g without virgule.
Figure 4. Iterations of MCMC results for progressive-stress model. Where ge-06 is g × 10 6 and g is value of parameter g without virgule.
Mathematics 10 01602 g004
Figure 5. Convergence of MCMC results for progressive-stress model.
Figure 5. Convergence of MCMC results for progressive-stress model.
Mathematics 10 01602 g005
Figure 6. Bayes point and interval estimates of the sub-survivor functions and the overall survivor function.
Figure 6. Bayes point and interval estimates of the sub-survivor functions and the overall survivor function.
Mathematics 10 01602 g006
Table 7. MLE and Kolmogorov–Smirnov test for each stress level of real data.
Table 7. MLE and Kolmogorov–Smirnov test for each stress level of real data.
First LevelSecond Level
EstimatesSEEstimatesSE
θ 3.38230.84292.60780.5590
a 1.10 × 10 5 3.96 × 10 6 8.33 × 10 6 3.08 × 10 6
b1.37560.03201.69940.0507
AIC554.7177493.5818
KSD0.11240.1113
p-value0.38520.4362
Table 8. MLEs and Bayesian estimates for parameters of progressive-stress model.
Table 8. MLEs and Bayesian estimates for parameters of progressive-stress model.
MLEBayesian
EstimatesSELowerUpperEstimatesSELowerUpper
θ11.97390.55900.000066.338812.38340.46914.042021.6353
a 1.84 × 10 5 3.08 × 10 6 1.84 × 10 5 1.84 × 10 5 1.92 × 10 5 2.75 × 10 6 1.08 × 10 6 1.71 × 10 5
b1.02600.05070.99791.05411.03220.04090.99971.0539
AIC1074.528
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Share and Cite

MDPI and ACS Style

Alotaibi, R.; Alamri, F.S.; Almetwally, E.M.; Wang, M.; Rezk, H. Classical and Bayesian Inference of a Progressive-Stress Model for the Nadarajah–Haghighi Distribution with Type II Progressive Censoring and Different Loss Functions. Mathematics 2022, 10, 1602. https://doi.org/10.3390/math10091602

AMA Style

Alotaibi R, Alamri FS, Almetwally EM, Wang M, Rezk H. Classical and Bayesian Inference of a Progressive-Stress Model for the Nadarajah–Haghighi Distribution with Type II Progressive Censoring and Different Loss Functions. Mathematics. 2022; 10(9):1602. https://doi.org/10.3390/math10091602

Chicago/Turabian Style

Alotaibi, Refah, Faten S. Alamri, Ehab M. Almetwally, Min Wang, and Hoda Rezk. 2022. "Classical and Bayesian Inference of a Progressive-Stress Model for the Nadarajah–Haghighi Distribution with Type II Progressive Censoring and Different Loss Functions" Mathematics 10, no. 9: 1602. https://doi.org/10.3390/math10091602

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

Article Metrics

Back to TopTop