Processing math: 100%
Research article

Estimation for inverse Weibull distribution under progressive type-Ⅱ censoring scheme

  • Received: 15 May 2023 Revised: 10 July 2023 Accepted: 11 July 2023 Published: 18 July 2023
  • MSC : 62F10, 62F15

  • This paper considers the statistical inferences of inverse Weibull distribution under progressive type-Ⅱ censored sample, which is a common distribution in reliability analysis. Two commonly used parameter estimation methods, maximum likelihood estimation and Bayesian estimation, are used in this paper, along with the inverse moment estimation. First, we derive the maximum likelihood estimators of parameters and propose Newtown-Raphson iteration method to solve these estimators. Assuming that shape and rate parameters are independent and follow gamma priors, we further obtain the Bayesian estimators by Lindley approximation. We also derive the inverse moment estimators and construct the generalized confidence intervals using the generalized pivotal quantity. To compare the estimation effects of these methods, we implement Monte Carlo simulation with the help of MATLAB. The simulation results show that the Bayesian estimation method outperforms the other two methods in terms of mean squared error. Finally, we verify the feasibility of these methods by analyzing a set of real data. The results indicate that the Bayesian estimation method provides more accurate estimates than the other two methods.

    Citation: Haiping Ren, Xue Hu. Estimation for inverse Weibull distribution under progressive type-Ⅱ censoring scheme[J]. AIMS Mathematics, 2023, 8(10): 22808-22829. doi: 10.3934/math.20231162

    Related Papers:

    [1] Najwan Alsadat, Mahmoud Abu-Moussa, Ali Sharawy . On the study of the recurrence relations and characterizations based on progressive first-failure censoring. AIMS Mathematics, 2024, 9(1): 481-494. doi: 10.3934/math.2024026
    [2] Xue Hu, Haiping Ren . Statistical inference of the stress-strength reliability for inverse Weibull distribution under an adaptive progressive type-Ⅱ censored sample. AIMS Mathematics, 2023, 8(12): 28465-28487. doi: 10.3934/math.20231457
    [3] Haiping Ren, Ziwen Zhang, Qin Gong . Estimation of Shannon entropy of the inverse exponential Rayleigh model under progressively Type-Ⅱ censored test. AIMS Mathematics, 2025, 10(4): 9378-9414. doi: 10.3934/math.2025434
    [4] Nora Nader, Dina A. Ramadan, Hanan Haj Ahmad, M. A. El-Damcese, B. S. El-Desouky . Optimizing analgesic pain relief time analysis through Bayesian and non-Bayesian approaches to new right truncated Fréchet-inverted Weibull distribution. AIMS Mathematics, 2023, 8(12): 31217-31245. doi: 10.3934/math.20231598
    [5] Samah M. Ahmed, Abdelfattah Mustafa . Estimation of the coefficients of variation for inverse power Lomax distribution. AIMS Mathematics, 2024, 9(12): 33423-33441. doi: 10.3934/math.20241595
    [6] M. Nagy, H. M. Barakat, M. A. Alawady, I. A. Husseiny, A. F. Alrasheedi, T. S. Taher, A. H. Mansi, M. O. Mohamed . Inference and other aspects for qWeibull distribution via generalized order statistics with applications to medical datasets. AIMS Mathematics, 2024, 9(4): 8311-8338. doi: 10.3934/math.2024404
    [7] Hassan Okasha, Mazen Nassar, Saeed A. Dobbah . E-Bayesian estimation of Burr Type XII model based on adaptive Type-Ⅱ progressive hybrid censored data. AIMS Mathematics, 2021, 6(4): 4173-4196. doi: 10.3934/math.2021247
    [8] Refah Alotaibi, Mazen Nassar, Zareen A. Khan, Ahmed Elshahhat . Analysis of Weibull progressively first-failure censored data with beta-binomial removals. AIMS Mathematics, 2024, 9(9): 24109-24142. doi: 10.3934/math.20241172
    [9] Ehab M. Almetwally, Ahlam H. Tolba, Dina A. Ramadan . Bayesian and non-Bayesian estimations for a flexible reduced logarithmic-inverse Lomax distribution under progressive hybrid type-Ⅰ censored data with a head and neck cancer application. AIMS Mathematics, 2025, 10(4): 9171-9201. doi: 10.3934/math.2025422
    [10] Refah Alotaibi, Mazen Nassar, Zareen A. Khan, Wejdan Ali Alajlan, Ahmed Elshahhat . Entropy evaluation in inverse Weibull unified hybrid censored data with application to mechanical components and head-neck cancer patients. AIMS Mathematics, 2025, 10(1): 1085-1115. doi: 10.3934/math.2025052
  • This paper considers the statistical inferences of inverse Weibull distribution under progressive type-Ⅱ censored sample, which is a common distribution in reliability analysis. Two commonly used parameter estimation methods, maximum likelihood estimation and Bayesian estimation, are used in this paper, along with the inverse moment estimation. First, we derive the maximum likelihood estimators of parameters and propose Newtown-Raphson iteration method to solve these estimators. Assuming that shape and rate parameters are independent and follow gamma priors, we further obtain the Bayesian estimators by Lindley approximation. We also derive the inverse moment estimators and construct the generalized confidence intervals using the generalized pivotal quantity. To compare the estimation effects of these methods, we implement Monte Carlo simulation with the help of MATLAB. The simulation results show that the Bayesian estimation method outperforms the other two methods in terms of mean squared error. Finally, we verify the feasibility of these methods by analyzing a set of real data. The results indicate that the Bayesian estimation method provides more accurate estimates than the other two methods.



    Kaller and Kamath [1] proposed two-parameter inverse Weibull distribution (IWD) to simulate the degradation of mechanical components of diesel engines. After that, IWD is considered an appropriate model to analyze lifetime data. For example, Abhijit and Anindya [2] found that IWD is superior to the normal distribution when using ultrasonic pulse velocity to measure concrete structures. Elio et al. [3] proposed a new model generated by appropriate mixing of IWD for modeling under extreme wind speed conditions. Langlands et al. [4] observed that breast cancer mortality data can be modeled and analyzed by IWD. Beyond that, IWD is widely used in reliability research. For example, Bi and Gui [5] considered the estimation of stress-strength reliability of IWD. They proposed an approximate maximum likelihood estimation for point and confidence interval estimations. Bayesian estimator and highest posterior density (HPD) confidence interval were derived using Gibbs sampling. Based on adaptive type-Ⅰ progressive hybrid censored scheme, Azm et al. [6] studied the estimation of unknown parameters of IWD when the data were competing risks data. The maximum likelihood estimation and Bayesian estimation were discussed. The asymptotic confidence intervals, the bootstrap confidence intervals and the HPD confidence intervals were derived. Then, two sets of real data were studied to illustrate maximum likelihood estimation and Bayesian estimation. Alslman and Helu [7] assumed that the two components were independent and identically distributed and considered the estimation of stress-strength reliability for IWD. Its estimators were derived by maximum likelihood estimation and maximum product of spacing method and compared using computer simulation. Shawky and Khan [8] assumed that stress and strength both followed IWD, and they focused on the multi-component stress-strength model. The estimation of reliability was obtained by maximum likelihood estimation. Monte Carlo simulation results indicate that the proposed estimating methods are effective.

    The probability density function (PDF), the cumulative distribution function (CDF) and the reliability function of IWD are respectively given by

    f(t;η,λ)=ηλtλ1eηtλ,t>0, η>0 , λ>0 (1.1)
    F(t;η,λ)=eηtλ,t>0, η>0 , λ>0 (1.2)

    and

    R(t)=1eηtλ. (1.3)

    Here, η is rate parameter and λ is shape parameter. For convenience, the PDF (1.1) of IWD will be denoted by IW(η,λ).

    The assessment of product reliability often relies on the collection of life data, which can be obtained through a life test. This test involves observing whether a group of test samples fail during the test and recording their corresponding failure time. If the life test continues until all the test samples fail, the failure time can be recorded for all the samples, resulting in complete data. However, if the test stops before all the test samples fail, the data collected are called censored data. With the continuous advancement of science and technology, products are becoming more reliable and have longer lifespans. Collecting complete data in a life test can be expensive, making reliability analysis based on censored data a popular research topic among scholars. A progressive type-Ⅱ censored sample can be expressed as follows: consider an experiment where n units are subjected to a life test at time zero, and the experimenter decides beforehand the number of failures to be observed, denoted by r. Upon observing the first failure time T1, Q1 out of the remaining n1 surviving units are randomly selected and removed. At the second observed failure time T2, Q2 out of the remaining n2Q1 surviving units are randomly selected and removed. This process continues until the r-th failure is observed at time Tr, and the remaining Qr=nrQ1Q2...Qr1 surviving units are all removed. The sample T=(T1,T2,...,Tr) is referred to as a progressively type-Ⅱ censored sample of size r from a sample of size n with censoring scheme Q=(Q1,Q2,...,Qr).

    IWD is one of the commonly used lifetime distributions in reliability estimation [5,6,7,8]. Most of the estimation methods are maximum likelihood estimation and Bayesian estimation, and there is a lack of research on inverse moment estimation. Therefore, this paper aims to provide three methods to compute the point estimations and construct generalized confidence intervals of unknown parameters.

    The rest of this paper is arranged as follows: In Section 2, the maximum likelihood estimators (MLEs) are obtained by Newtown-Raphson method. The Lindley approximation is proposed to derive the Bayesian estimators (BEs) in Section 3. The inverse moment estimators (IMEs) and the construction of the generalized confidence intervals (GCIs) are discussed in Section 4. In Section 5, Monte Carlo simulation is conducted to evaluate the effect of these methods. Section 6 gives a set of real data as a demonstration. Finally, Section 7 gives the conclusions of this paper.

    In this section, we will discuss the MLEs of η, λ and R(t). Due to the nonlinear nature of the likelihood equations, the Newton-Raphson method is considered to solve likelihood equations numerically.

    Let T=(T1,T2,...,Tr) be the collected progressive type-Ⅱ sample under the censoring scheme Q=(Q1,Q2,...,Qr). Denote t=(t1,t2,...,tr) as the observation of T. We can easily get the likelihood function l(η,λ;t) as follows:

    l(η,λ;t)=δηrλrri=1tλ1ieηtλi(1eηtλi)Qi, (2.1)

    where δ=n(nQ11)(nQ1Q22)...(nQ1Q2...Qr1r+1).

    Then the log-likelihood function L(η,λ;t) is

    L(η,λ;t)=lnl(η,λ;t)=lnδ+rlnηλri=1[(λ+1)lnti+ηtλiQiln(1eηtλi)]. (2.2)

    Thus, the partial derivatives of L(η,λ;t) with respect to η and λ are respectively given by

    L(η,λ;t)η=rηri=1[tλiQitλiexp(ηtλi)1exp(ηtλi)], (2.3)
    L(η,λ;t)λ=rλri=1[lntiηtλilnti+ηQitλiexp(ηtλi)lnti1exp(ηtλi)]. (2.4)

    Denote the MLEs of η and λ as ˆηML and ˆλML respectively, and they are the solutions of likelihood equations (2.5). According to the invariance of the maximum likelihood estimation, the MLE ˆRML(t) is obtained by replacing the parameters with ˆηML and ˆλML. Since (2.3) and (2.4) are non-linear, the Newtown-Raphson method is considered to solve likelihood equations numerically. The elements of the Jacobi matrix are given from (2.6) to (2.9).

    {L(η,λ;t)η=0L(η,λ;t)λ=0 (2.5)
    2L(η,λ;t)η2=rη2ri=1Qit2λiexp(ηtλi)[1exp(ηtλi)]2 (2.6)
    2L(η,λ;t)ηλ=ri=1[tλilnti+Qitλieηtλi(lnti)(eηtλi+ηtλi1)(1eηtλi)2] (2.7)
    2L(η,λ;t)λη=ri=1[tλilnti+Qitλieηtλi(lnti)(eηtλi+ηtλi1)(1eηtλi)2] (2.8)
    2L(η,λ;t)λ2=rλ2ηri=1[tλi(lnti)2Qi(lnti)2tλieηtλi(eηtλi+ηtλi1)(1eηtλi)2]. (2.9)

    Define

    H(θ)=[L(η,λ;t)ηL(η,λ;t)λ]

    and

    H(θ)=[2L(η,λ;t)η22L(η,λ;t)ηλ2L(η,λ;t)λη2L(η,λ;t)λ2],

    where θ=(η,λ). The steps involved in Newton-Raphson iteration method for obtaining the MLEs of η and λ are given below.

    Step 1. Pick an arbitrary starting estimate θ0, and desired precision ε=105.

    Step 2. Update θ0 as θnew=θ0[H(θ0)]1H(θ0), where [H(θ0)]1 is the inverse matrix of H(θ0).

    Step 3. If |θnewθ0|ε, then ˆθML=θnew.

    Step 4. If |θnewθ0|>ε, set θ0=θnew and return to Step 2.

    Step 5. Repeat from Step 2 to Step 4 until the condition in Step 3 is achieved.

    Statistical inference requires three kinds of information: overall information, sample information and prior information. The statistical inference based on the first two kinds of information is called classical statistics, and the statistical inference based on comprehensive consideration of the three kinds of information is called Bayes statistics. Prior information already exists before sampling, which mostly comes from experience and historical data. The distribution obtained by processing prior information is called a prior distribution. We all know that a random variable can be described by a certain distribution. Bayesian scholars believe that an unknown parameter can be regarded as a random variable. In other words, an unknown parameter can also be described by a certain distribution, that is, a prior distribution. Kundu and Howlader [9] considered the estimation of parameters of IWD using the Bayesian approach under the squared error loss function when the sample is a type-Ⅱ censoring sample. Akgul et al. [10] discussed Bayes estimation of the step-stress partially accelerated life test model with type-Ⅰ censored sample for the IWD. The point estimators were obtained using the Lindley approximation and Tierney–Kadane approximation. The credible intervals were constructed using the Gibbs sampling method. Helu and Samawi [11] considered the Bayesian inferences based on IWD based on progressive first-failure censored data. The point estimators were derived under three loss functions. Additionally, the estimators were calculated by Lindley approximation. Based on generalized adaptive progressively hybrid censored sample, Lee [12] considered the estimation of uncertainty measure for IWD. For the BE and HPD confidence interval, the Tierney-Kadane approximation and importance sampling technique were proposed.

    In this section, the BEs of η, λ and R(t) are derived under the symmetric entropy (SE) loss function, scale squared error (SSE) loss function and LINEX loss function. Since there is a complex ratio of two integrals in BEs, Lindley approximation is proposed to solve this problem.

    (i) The SE loss function is defined as (Xu et al. [13]):

    S1(θ,ˆθ)=θˆθ+ˆθθ2. (3.1)

    (ii) The SSE loss function is defined as (Song et al. [14]):

    S2(θ,ˆθ)=(θˆθ)2θd. (3.2)

    (iii) The LINEX loss function is defined as (Varian [15]):

    S3(θ,ˆθ)=ea(ˆθθ)a(ˆθθ)1 , a>0, (3.3)

    where ˆθ is the estimator of unknown parameter θ and d is a nonnegative integer.

    As an important part of Bayesian estimation, the selection of prior distribution will directly affect the final Bayesian estimation. It usually follows two rules making full use of prior information, such as empirical and historical data and being convenient for computational use. The most widely used prior distributions are mainly non-informative prior distribution, conjugate prior distribution and hierarchical prior distribution. The gamma prior belongs to the conjugate prior, which makes the computation process of estimation results easier.

    Assume the prior distributions of η and λ are gamma prior. η follows Gamma(a1,b1) and λ follows Gamma(a2,b2).

    π(η)ηb11ea1η a1>0, b1>0. (3.4)
    π(λ)λb21ea2λ a2>0, b2>0. (3.5)

    Based on (3.4) and (3.5), the joint prior distribution is

    π(η,λ)ηb11λb21ea1ηa2λ. (3.6)

    Using (2.1) and (3.6), the posterior distribution is

    π(η,λ|T)=Kηr+b11λr+b21ea1ηa2λri=1tλ1ieηtλi(1eηtλi)Qi, (3.7)

    where K=[+0+0ηr+b11λr+b21ea1ηa2λri=1tλ1ieηtλi(1eηtλi)Qidηdλ]1.

    Based on (3.7), the margin posterior distribution of η is

    π(η|T)=Kηr+b11ea1η+0[λr+b21ea2λri=1tλ1ieηtλi(1eηtλi)Qi]dλ. (3.8)

    The margin posterior distribution of λ is

    π(λ|T)=Kλr+b21ea2λ+0[ηr+b11ea1ηri=1tλ1ieηtλi(1eηtλi)Qi]dη. (3.9)

    Lindley [16] proposed an approximation algorithm to calculate the ratio of two integrals in the form:

    I(t)=E[W(η,λ)|T]=W(η,λ)eL(η,λ;t)+J(η,λ)d(η,λ)eL(η,λ;t)+J(η,λ)d(η,λ). (3.10)

    Here is a continuous function in η and λ, L(η,λ;t) as shown in (2.2) and J(η,λ) is the logarithm of joint prior distribution (3.6). This ratio usually occurs in BE, which is why the Lindley approximation is often used to calculate the Bayesian estimator.

    The expression (3.10) can be approximated by (3.11) under regularity conditions or with a large sample size. Here A, B and C are given by (3.12). Lηηη denotes the third derivative of log-likelihood function (2.2) for η, and ˆLηηη represents the value of Lηηη at η=ˆηML. ψij is the element of inverse matrix of Lij (i,j=η,λ), and ˆψij represents the value of ψij at η=ˆηML and λ=ˆλML. Other terms are defined as the same as the above rules. The detailed expressions are presented in (3.13) to (3.18).

    I(t)=W(ˆηML,ˆλML)+12(A+B+C) (3.11)
    A=(ˆWηη+2ˆWηˆJη)ˆψηη+(ˆWλη+2ˆWλˆJη)ˆψλη+(ˆWηλ+2ˆWηˆJλ)ˆψηλ+(ˆWλλ+2ˆWλˆJλ)ˆψλλB=(ˆWηˆψηη+ˆWλˆψηλ)(ˆLηηηˆψηη+ˆLηληˆψηλ+ˆLληηˆψλη+ˆLλληˆψλλ)C=(ˆWηˆψλη+ˆWλˆψλλ)(ˆLηηλˆψηη + ˆLηλλˆψηλ+ˆLληλˆψλη+ˆLλλλˆψλλ) (3.12)
    Lηηη=2rη3+ri=1Qit3λieηtλi(1+eηtλi)(1eηtλi)3 (3.13)
    Lηηλ=2ri=1Qit2λieηtλilnti(1eηtλi)2ηri=1Qit3λieηtλi(lnti)(1+eηtλi)(1eηtλi)3 (3.14)
    Lληλ=ri=1tλi(lnti)2+η2ri=1Qit3λieηtλi(lnti)2(1+eηtλi)(1eηtλi)33ηri=1Qit2λieηtλi(lnti)2(1eηtλi)2+ri=1Qitλieηtλi(lnti)21eηtλi (3.15)
    Lλλλ=2rλ3+ηri=1tλi(lnti)3η3ri=1Qit3λieηtλi(lnti)3(1+eηtλi)(1eηtλi)3+η2ri=1Qit2λieηtλi(lnti)3(1eηtλi)2+ηri=1Qitλieηtλi(lnti)3(1+eηtλi)(1eηtλi)2 (3.16)
    Lηλη=Lληη=Lηηλ , Lλλη=Lηλλ=Lληλ (3.17)
    Jη=b11ηa1 , Jλ=b21λa2. (3.18)

    Lemma 1. Suppose that T is a random sample. The BE ˆθSE of unknown parameter θ under the SE loss function (3.1) for any prior distribution π(θ) is

    ˆθSE=[E(θ|T)E(θ1|T)]12, (3.19)

    where E(θ|T) and E(θ1|T) denote the posterior expectations of θ and θ1.

    Proof. Based on the SE loss function (3.1), the Bayesian risk of ˆθSE is

    R(ˆθSE)=Eθ(E(S1(θ,ˆθSE)|T)).

    To minimize R(ˆθSE), only need to minimize E(S1(θ,ˆθSE)|T). Denote h1(ˆθSE)=E(S1(θ,ˆθSE)|T) for convenience.

    Because

    h1(ˆθSE)=ˆθSE1E(θ|T)+ˆθSEE(θ1|T)2,

    and the derivative is

    h1(ˆθSE)=ˆθSE2E(θ|T)+E(θ1|T).

    The BE ˆθSE can be obtained by solving the equation h1(ˆθSE)=0.

    According to Lemma 1, the BEs ˆηSE, ˆλSE and ˆRSE(t) under the SE loss function can be written as:

    ˆηSE=[E(η|T)E(η1|T)]12, (3.20)
    ˆλSE=[E(λ|T)E(λ1|T)]12, (3.21)

    and

    ˆRSE(t)=[E(R(t)|T)E(R1(t)|T)]12. (3.22)

    From the marginal posterior distribution (3.8), the BE (3.20) may be written as

    ˆηSE=[+0ηπ(η|T)dη+0η1π(η|T)dη]12=[+0+0ηr+b1λr+b21ea1ηa2λri=1tλ1ieηtλi(1eηtλi)Qidλdη+0+0ηr+b12λr+b21ea1ηa2λri=1tλ1ieηtλi(1eηtλi)Qidλdη]12. (3.23)

    From the marginal posterior distribution (3.9), the BE (3.21) may be written as

    ˆλSE=[+0λπ(λ|T)dλ+0λ1π(λ|T)dλ]12=[+0+0λr+b2ηr+b11ea1ηa2λri=1tλ1ieηtλi(1eηtλi)Qidηdλ+0+0λr+b22ηr+b11ea1ηa2λri=1tλ1ieηtλi(1eηtλi)Qidηdλ]12. (3.24)

    From the posterior distribution (3.7), the BE (3.22) may be written as

    ˆRSE(t)=[+0+0(1eηtλ)π(η,λ|T)dηdλ+0+0(1eηtλ)1π(η,λ|T)dηdλ]12=[+0+0(1eηtλ)ηr+b11λr+b21ea1ηa2λri=1tλ1ieηtλi(1eηtλi)Qidηdλ+0+0(1eηtλ)1ηr+b11λr+b21ea1ηa2λri=1tλ1ieηtλi(1eηtλi)Qidηdλ]12. (3.25)

    It is obvious that (3.23)–(3.25) cannot be evaluated explicitly. Thus, the Lindley approximation is used to approximate them.

    (i) When W(η,λ)=η, there are

    Wη=1 , Wλ=Wηη=Wηλ=Wλη=Wλλ=0. (3.26)

    Submitting W(η,λ)=η and (3.26) in the expression (3.11), the posterior expectation E(η|T) may be written as

    E(η|T)=ˆηML+ˆJηˆψηη+ˆJλˆψηλ+12[ˆψηη(ˆLηηηˆψηη+ˆLηληˆψηλ+ˆLληηˆψλη+ˆLλληˆψλλ)+ˆψλη(ˆLηηλˆψηη + ˆLηλλˆψηλ+ˆLληλˆψλη+ˆLλλλˆψλλ)]. (3.27)

    (ii) When W(η,λ)=η1, there are

    Wη=1η2 , Wηη=2η3 , Wλ=Wηλ=Wλη=Wλλ=0. (3.28)

    Submitting W(η,λ)=η1 and (3.28) in the expression (3.11), the posterior expectation E(η1|T) may be written as

    E(η1|T)=12ˆη2ML[ˆψηη(ˆLηηηˆψηη+ˆLηληˆψηλ+ˆLληηˆψλη+ˆLλληˆψλλ)ˆψλη(ˆLηηλˆψηη + ˆLηλλˆψηλ+ˆLληλˆψλη+ˆLλλλˆψλλ)]+ˆη1ML+(ˆη3MLˆη2MLˆJη)ˆψηηˆη2MLˆJλˆψηλ. (3.29)

    The BE ˆηSE of rate parameter η under the SE loss function can be obtained by submitting (3.27) and (3.29) in the expression (3.20).

    Using Lindley approximation, the BEs ˆλSE and ˆRSE(t) under SE loss function are obtained as similar to the above steps.

    Lemma2. Suppose that T is a set of simple random samples. The BE ˆθSSE of unknown parameter θ under the SSE loss function (3.2) for any prior distribution π(θ) is

    ˆθSSE=E(θ1d|T)E(θd|T). (3.30)

    Proof. The Bayesian risk of ˆθSSE based on SSE loss function (3.2) is

    R(ˆθSSE)=Eθ(E(S2(θ,ˆθSSE)|T)).

    Denote h2(ˆθSSE)=E(S2(θ,ˆθSSE)|T), and

    h2(ˆθSSE)=E(θ2d|T)2ˆθSSEE(θ1d|T)+ˆθ2SSEE(θd|T).

    The derivative of h2(ˆθSSE) is

    h2(ˆθSSE)=2ˆθSSEE(θd|T)2E(θ1d|T).

    Therefore, the BE ˆθSSE under the SSE loss function is derived by solving equation h2(ˆθSSE)=0.

    According to Lemma 2, the BEs ˆηSSE, ˆλSSE and ˆRSSE(t) under SSE loss function are presented in (3.31) and (3.33) respectively.

    ˆηSSE=E(η1d|T)E(ηd|T), (3.31)
    ˆλSSE=E(λ1d|T)E(λd|T), (3.32)

    and

    ˆRSSE(t)=E[R1d(t)|T]E[Rd(t)|T]. (3.33)

    From the marginal posterior distribution (3.8), the BE (3.31) may be written as

    ˆηSSE=+0η1dπ(η|T)dη+0ηdπ(η|T)dη=+0+0ηr+b1dλr+b21ea1ηa2λri=1tλ1ieηtλi(1eηtλi)Qidλdη+0+0ηr+b1d1λr+b21ea1ηa2λri=1tλ1ieηtλi(1eηtλi)Qidλdη. (3.34)

    From the marginal posterior distribution (3.9), the BE (3.32) can be written as

    ˆλSSE=+0λ1dπ(λ|T)dλ+0λdπ(λ|T)dλ=+0+0λr+b2dηr+b11ea1ηa2λri=1tλ1ieηtλi(1eηtλi)Qidηdλ+0+0λr+b21dηr+b11ea1ηa2λri=1tλ1ieηtλi(1eηtλi)Qidηdλ. (3.35)

    From the posterior distribution (3.7), the BE (3.33) can be written as

    ˆRSSE(t)=+0+0R1d(t)π(η,λ|T)dηdλ+0+0Rd(t)π(η,λ|T)dηdλ=+0+0(1eηtλ)1dηr+b11λr+b21ea1ηa2λri=1tλ1ieηtλi(1eηtλi)Qidηdλ+0+0(1eηtλ)dηr+b11λr+b21ea1ηa2λri=1tλ1ieηtλi(1eηtλi)Qidηdλ. (3.36)

    Then, Lindley approximation is used to approximate (3.34) to (3.36).

    (i) When W(η,λ)=η1d, there are

    Wη=(1d)ηd , Wηη=d(1d)ηd1 , Wλ=Wηλ=Wλη=Wλλ=0. (3.37)

    Putting W(η,λ)=η1d and (3.37) into (3.11), the posterior expectation E(η1d|T) can be written as

    E(η1d|T)=ˆη1dML+(1d)ˆηdMLˆJηˆψηη+(1d)ˆηdMLˆJλˆψηλ+12[d(1d)ˆηd1MLˆψηη+(1d)ˆηdMLˆψηη(ˆLηηηˆψηη+ˆLηληˆψηλ+ˆLληηˆψλη+ˆLλλαˆψλλ)+(1d)ˆηdMLˆψλη(ˆLηηλˆψηη + ˆLηλλˆψηλ+ˆLληλˆψλη+ˆLλλλˆψλλ)]. (3.38)

    (ii) When W(η,λ)=ηd, there are

    Wη=dηd1 , Wηη=d(d+1)ηd2 , Wλ=Wηλ=Wλη=Wλλ=0. (3.39)

    Putting W(η,λ)=ηd and (3.39) into (3.11), the posterior expectation E(ηd|T) can be written as

    E(ηd|T)=ˆηdMLdˆηd1MLˆJηˆψηηdˆηd1MLˆJλˆψηλ+12[d(d+1)ˆηd2MLˆψηηdˆηd1MLˆψηη(ˆLηηηˆψηη+ˆLηληˆψηλ+ˆLληηˆψλη+ˆLλληˆψλλ)dˆηd1MLˆψλη(ˆLηηλˆψηη + ˆLηλλˆψηλ+ˆLληλˆψλη+ˆLλλλˆψλλ)]. (3.40)

    Hence the BE ˆηSSE under the SSE loss function are obtained by submitting (3.38) and (3.40) in (3.31).

    Using Lindley approximation, the BEs ˆλSSE and ˆRSSE(t) under SSE loss function can be obtained by the similar steps.

    Lemma 3. Suppose that T is a set of simple random samples. The BE ˆθL of unknown parameter θ under the LINEX loss function (3.3) for any prior distribution π(θ) is

    ˆθL=1aln[E(eaθ|T)]. (3.41)

    Proof. The Bayesian risk of ˆθL based on LINEX loss function (3.3) is

    R(ˆθL)=Eθ(E(S3(θ,ˆθL)|T)).

    Denote h3(ˆθL)=E(S3(θ,ˆθL)|T), and

    h3(ˆθL)=E[ea(ˆθLθ)|T]aˆθL+aE(θ|T).

    The derivative of h3(ˆθL) is

    h3(ˆθL)=aea(ˆθLθ)E[ea(ˆθLθ)|T]a.

    Therefore, the BE ˆθL under the LINEX loss function is derived by solving equation h3(ˆθL)=0.

    It follows from Lemma 3 that the BEs under LINEX loss function are

    ˆηL=1aln[E(eaη|T)] (3.42)
    ˆλL=1aln[E(eaλ|T)] (3.43)
    ˆRL(t)=1aln[E(exp(aeηtλ)|T)]. (3.44)

    Subsequently, (3.42)–(3.44) can be written as

    ˆηL=1aln[+0eaηπ(η|T)dη]=1aln[K+0+0ηr+b11λr+b21ea1ηa2λaηri=1tλ1ieηtλi(1eηtλi)Qidηdλ] (3.45)
    ˆλL=1aln[+0eaλπ(λ|T)dλ]=1aln[K+0+0λr+b21ηr+b11ea1ηa2λaλri=1tλ1ieηtλi(1eηtλi)Qidηdλ] (3.46)
    ˆRL(t)=1aln[+0+0exp(aeηtλ)π(η,λ|T)dηdλ]=1aln[K+0+0exp(aeηtλ)ηr+b11λr+b21ea1ηa2λri=1tλ1ieηtλi(1eηtλi)Qidηdλ]. (3.47)

    Next, the explicit expressions of these BEs are obtained by using the Lindley approximation. When W(η,λ)=eaη, there are

    Wη=aeaη, Wηη=a2eaη, Wλ=Wηλ=Wλη=Wλλ=0. (3.48)

    According to Lindley's formula (3.11), the posterior expectation E(eaη|T) can be written as

    E(eaη|T)=eaˆηML+12[(a2eaˆηML2aeaˆηMLˆJη)ˆψηη2aeaˆηMLˆJλˆψηλaeaˆηMLˆψηη(ˆLηηηˆψηη+ˆLηληˆψηλ+ˆLληηˆψλη+ˆLλληˆψλλ)aeaˆηMLˆψλη(ˆLηηλˆψηη + ˆLηλλˆψηλ+ˆLληλˆψλη+ˆLλλλˆψλλ)]. (3.49)

    The BE ˆηL under LINEX loss function is derived by substituting (3.49) into (3.42). The BEs ˆλL and ˆRL(t) can be acquired using a comparable method to the aforementioned steps, and therefore will not be reiterated here.

    In Sections 2 and 3, the MLEs and BEs have been derived. However, we cannot obtain the explicit forms of MLEs and BEs easily by using both methods. In addition, it is necessary to select the appropriate initial values when using the Newtown-Raphson iteration method. Therefore, the generalized pivot quantity is constructed for deriving IMEs and GCIs in this Section. Compared with the maximum likelihood estimation and Bayes estimation, the inverse moment estimation is much simpler in calculation. It only needs some mathematical transformations and finally solves the equations. Wang [17] proposed a new method and named it as inverse moment estimation method in 1992. Additionally, the method was applied to parameter estimation of Weibull distribution. After that, inverse moment estimation has been widely used and studied. For example, Luo et al. [18] used the inverse third-moment method when forecasting a single time series using a large number of predictors in the presence of a possible nonlinear forecast function. Qin and Yuan [19] proposed an ensemble of IMEs to explore the central subspace. Based on progressive censored data, Gao et al. [20] proposed the pivotal inference for inverse exponentiated Rayleigh distribution. The point estimators were derived using the method that combined pivotal quantity with inverse moment estimation.

    In this section, the IMEs and GCIs of η, λ and R(t) are obtained by constructing the generalized pivot quantity.

    Definition 1. Assume that T is a random variable and t is the observation of T, and θ1 is an interest parameter and θ2 is a nuisance parameter. A function G(T;t,θ1,θ2) is called a generalized pivotal quantity if it satisfies the following conditions:

    (1) Given t, the distribution of G(T;t,θ1,θ2) is unrelated to both θ1 and θ2.

    (2) Given t, the observation G(t;t,θ1,θ2) of generalized pivotal quantity G(T;t,θ1,θ2) is unrelated to θ2.

    First, let

    Xi=ηTλi, i=1,2,...,r. (4.1)

    The distribution of Xi is

    FX(xi)=P(Xixi)=P(ηTλixi)=1exi. (4.2)

    Let Exp(1) be the standard exponential distribution. It is obvious that XiExp(1), and Xr<Xr1<...<X1. Let

    {S1=rXrS2=(r1)(Xr1Xr)S3=(r2)(Xr2Xr1)...Sr=X1X2. (4.3)

    Then, S1,S2,...,Sr are independent of each other and SiExp(1). Denote U=2ri=1Si and V=2S1, so U follows χ2 distribution with 2r2 degrees of freedom and V follows χ2 distribution with 2 degrees of freedom. Finally, let

    G1=V2/V2U2r2U2r2=r(r1)Tλrri=1TλirTλr (4.4)
    G2=U+V=2ηri=1Tλi. (4.5)

    Therefore, G1 and G2 are independent, G1 follows F distribution with 2 and 2r2 degrees of freedom and G2 follows χ2 distribution with 2r degrees of freedom. According to Definition 1, G1 is the generalized pivotal quantity of λ, but G2 is neither a generalized pivotal quantity of η nor λ. Since (r1)(r2)1 is the mean of F(2,2r2) and 2r is the mean of χ2(2r), Theorem 1 can be derived.

    Theorem 1. Let T=(T1,T2,...,Tr) be a progressive type-Ⅱ censored sample following IW(ηλ), and let Q=(Q1,Q2,...,Qr) be the censoring scheme. Denote t=(t1,t2,...,tr) as the observation of T. The IME ˆηG of η and the IME ˆλG of λ are determined by the following expressions. The IME ˆRG(t) is obtained by replacing the parameters with ˆηG and ˆλG.

    {G1=r(r1)tλrri=1tλirtλr=r1r2G2=2ηri=1tˆλGi=2r. (4.6)

    This section will discuss the GCIs by generalized pivotal quantity.

    Lemma 3. Suppose that a set of constants ki (i=1,2,...,r) satisfy 0<k1<k2<...<kr and denote G(λ)=r(r1)kλrri=1kλirkλr.

    (i) G(λ) decreases monotonically when λ>0;

    (ii) The equation G(λ)=k has only one solution, which k>0 and k is a constant.

    Proof. (i) The derivative of G(λ) is

    G(λ)=r(r1)kλr[ri=1kλilnki(lnkr)ri=1kλi]=r(r1)kλr(k1lnk1+k2lnk2+...+krlnkrk1lnkrk2lnkr...krlnkr)=r(r1)kλr[k1(lnk1lnkr)+k2(lnk2lnkr)+...+kr1(lnkr1lnkr)].

    According to 0<k1<k2<...<kr, it can be derived 0<lnk1<lnk2<...<lnkr. That is lnkilnkr<0 (i=1,2,...,r1). Therefore, G(λ) decreases monotonically when λ>0.

    (ii) Suppose that the equation G(λ)=k has two unequal solutions, λ1 and λ2 respectively. Based on G(λ1)=G(λ2), there is

    r(r1)kλ1rri=1kλ1irkλ1r=r(r1)kλ2rri=1kλ2irkλ2r.

    That is

    1ri=1(kikr)λ1r=1ri=1(kikr)λ2r.

    Here, (kikr)λ is monotone because of kikr1. Hence the above expression is inconsistent with the supposition, and the equation G(λ)=k has only one solution.

    Theorem 2. Let T=(T1,T2,...,Tr) with observation t=(t1,t2,...,tr) be a progressive type-Ⅱ censored sample following IW(ηλ), and let Q=(Q1,Q2,...,Qr) be the censoring scheme. Fω(2,2r2) and χ2ω(2r) denote the upper quantile of F(2,2r2) and χ2(2r) respectively with ω=11γ2. The 100(1γ)% GCIs of η and λ are given as follows:

    {φ(t1,...,tr,F1ω(2,2r2))<λ<φ(t1,...,tr,Fω(2,2r2))(2ri=1tˆλGi)1χ2ω(2r)<η<(2ri=1tˆλGi)1χ21ω(2r) (4.7)

    where φ(t1,...,tr,F1ω(2,2r2)) is the solution of equation (4.8) and φ(t1,...,tr,Fω(2,2r2)) is the solution of Eq (4.9).

    r(r1)tλrri=1tλirtλr=F1ω(2,2r2) (4.8)
    r(r1)tλrri=1tλirtλr=Fω(2,2r2). (4.9)

    Proof. From Section 4.1, there are G1 F(2,2r2) and G2χ2(2r). G1 and G2 are independent, so

    P(F1ω(2,2r2)<G1<Fω(2,2r2),χ2ω(2r)<G2<χ21ω(2r))=P(F1ω(2,2r2)<G1<Fω(2,2r2))P(χ2ω(2r)<G2<χ21ω(2r))=1γ1γ=1γ.

    The F1ω(2,2r2)<G1<Fω(2,2r2) may be written as

    F1ω(2,2r2)<r(r1)tλrri=1tλirtλr<Fω(2,2r2).

    According to Lemma 3, F1ω(2,2r2) and Fω(2,2r2) are positive constants, the above inequation is equivalent to

    φ(t1,...,tr,F1ω(2,2r2))<λ<φ(t1,...,tr,Fω(2,2r2)).

    χ2ω(2r)<G2<χ21ω(2r) is equivalent to

    (2ri=1tˆλGi)1χ2ω(2r)<η<(2ri=1tˆλGi)1χ21ω(2r).

    In this Section, the proposed estimation methods are compared using MATLAB. For point estimates, the mean squared errors (MSEs) are calculated by Eq (5.1). For interval estimates, coverage probability (CP) is used to reflect the performance of GCIs. The simulation is carried out under true value (ηreal,λreal)=(2,2) and different n, r and Q, and the trials are N at 1000 times. The hyper-parameter of the prior distribution is (a1,b1)=(2,1.8) and (a2,b2)=(1.5,2), and the parameters of SSE loss function and LINEX loss function are d=4 and a=4 respectively. The MSEs of η, and λ are shown in Tables 1 and 2 respectively, and the MSEs of R(t) are shown in Table 3 with t=3. The CP values is shown in Table 4.

    Table 1.  The MSEs of η.
    n r Q MSE
    ˆηML ˆηSE ˆηSSE ˆηL ˆηG
    20 10 (3*1, 0*4, 7*1, 0*4) 0.4045 0.1702 0.2618 0.3882 0.4207
    (10*1, 0*9) 0.6676 0.1367 0.3823 0.4630 0.5111
    (2*5, 0*5) 0.5663 0.1397 0.4627 0.2966 0.3729
    20 (0*20) 0.3788 0.1225 0.2355 0.2871 0.2957
    30 10 (5*3, 0*2, 5*1, 0*4) 0.3418 0.1099 0.2446 0.2133 0.5184
    (20*1, 0*9) 0.3413 0.1399 0.3037 0.2509 0.2788
    (4*5, 0*5) 0.6002 0.1364 0.3790 0.2189 0.5100
    20 (0*10, 2*5, 0*5) 0.2588 0.0913 0.1663 0.1542 0.3119
    (10*1, 0*19) 0.2918 0.1106 0.2028 0.2170 0.1997
    (1*10, 0*10) 0.2529 0.1004 0.1790 0.1629 0.2052
    30 (0*30) 0.1962 0.0944 0.1454 0.1435 0.1687
    50 15 (2*4, 1*6, 0*5) 0.1479 0.0860 0.1356 0.1850 0.3282
    (35*1, 0*14) 0.2021 0.1098 0.1999 0.1703 0.1994
    (7*5, 0*10) 0.1516 0.0886 0.1450 0.1474 0.4038
    20 (5*6, 0*10, 0*4) 0.1304 0.0784 0.1153 0.1297 0.2861
    (30*1, 0*19) 0.1854 0.0969 0.1595 0.1506 0.1690
    (6*5, 0*15) 0.1316 0.0858 0.1324 0.1273 0.2458
    30 (3*4, 0*7, 1*8, 0*11) 0.1121 0.0689 0.0919 0.0975 0.2010
    (4*5, 0*25) 0.1270 0.0811 0.1137 0.1012 0.1412
    (20*1, 0*29) 0.1578 0.0916 0.1315 0.1108 0.1525
    50 (0*50) 0.1038 0.0667 0.0844 0.0903 0.1042

     | Show Table
    DownLoad: CSV
    Table 2.  The MSEs of λ.
    n r Q MSE
    ˆλML ˆλSE ˆλSSE ˆλL ˆλG
    20 10 (3*1, 0*4, 7*1, 0*4) 0.4838 0.2817 0.2544 0.2704 0.5386
    (10*1, 0*9) 0.5354 0.1750 0.2800 0.3230 0.4666
    (2*5, 0*5) 0.4809 0.1760 0.2538 0.3093 0.5099
    20 (0*20) 0.1701 0.0966 0.1241 0.1230 0.3621
    30 10 (5*3, 0*2, 5*1, 0*4) 0.4990 0.1847 0.2592 0.2567 0.5616
    (20*1, 0*9) 0.4024 0.1763 0.2308 0.2739 0.4615
    (4*5, 0*5) 0.5383 0.1622 0.2823 0.2722 0.5173
    20 (0*10, 2*5, 0*5) 0.1760 0.0967 0.1166 0.1277 0.4387
    (10*1, 0*19) 0.1746 0.0979 0.1198 0.1160 0.3448
    (1*10, 0*10) 0.1874 0.1099 0.1303 0.1148 0.3156
    30 (0*30) 0.0982 0.0670 0.0800 0.0774 0.2799
    50 15 (2*4, 1*6, 0*5) 0.2545 0.1383 0.1596 0.1655 0.4156
    (35*1, 0*14) 0.1876 0.1183 0.1280 0.1448 0.3584
    (7*5, 0*10) 0.2201 0.1309 0.1416 0.1609 0.3567
    20 (5*6, 0*10, 0*4) 0.1587 0.1010 0.1093 0.1194 0.3233
    (30*1, 0*19) 0.1461 0.1004 0.1067 0.0938 0.3066
    (6*5, 0*15) 0.1500 0.1023 0.1121 0.1058 0.2882
    30 (3*4, 0*7, 1*8, 0*11) 0.1099 0.0760 0.0825 0.0848 0.2882
    (4*5, 0*25) 0.0946 0.0700 0.0761 0.0719 0.2754
    (20*1, 0*29) 0.0904 0.0672 0.0735 0.0728 0.2723
    50 (0*50) 0.0575 0.0447 0.0476 0.0488 0.2112

     | Show Table
    DownLoad: CSV
    Table 3.  The MSEs of R(t).
    n r Q MSE
    ˆRML(t) ˆRSE(t) ˆRSSE(t) ˆRL(t) ˆRG(t)
    20 10 (3*1, 0*4, 7*1, 0*4) 0.0088 0.0061 0.0113 0.0110 0.0129
    (10*1, 0*9) 0.0107 0.0069 0.0121 0.0129 0.0166
    (2*5, 0*5) 0.0090 0.0021 0.0111 0.0108 0.0142
    20 (0*20) 0.0052 0.0040 0.0096 0.0913 0.0128
    30 10 (5*3, 0*2, 5*1, 0*4) 0.0092 0.0064 0.0103 0.0129 0.0131
    (20*1, 0*9) 0.0091 0.0064 0.0123 0.0180 0.0166
    (4*5, 0*5) 0.0112 0.0080 0.0096 0.0129 0.0120
    20 (0*10, 2*5, 0*5) 0.0047 0.0037 0.0061 0.0119 0.0082
    (10*1, 0*19) 0.0049 0.0039 0.0058 0.0152 0.0122
    (1*10, 0*10) 0.0048 0.0038 0.0057 0.0081 0.0095
    30 (0*30) 0.0036 0.0030 0.0041 0.0068 0.0101
    50 15 (2*4, 1*6, 0*5) 0.0059 0.0045 0.0070 0.0066 0.0099
    (35*1, 0*14) 0.0063 0.0050 0.0078 0.0104 0.0127
    (7*5, 0*10) 0.0064 0.0051 0.0083 0.0083 0.0100
    20 (5*6, 0*10, 0*4) 0.0044 0.0036 0.0056 0.0079 0.0092
    (30*1, 0*19) 0.0053 0.0043 0.0062 0.0070 0.0111
    (6*5, 0*15) 0.0053 0.0044 0.0064 0.0088 0.0096
    30 (3*4, 0*7, 1*8, 0*11) 0.0034 0.0029 0.0039 0.0042 0.0072
    (4*5, 0*25) 0.0033 0.0029 0.0039 0.0045 0.0087
    (20*1, 0*29) 0.0032 0.0027 0.0037 0.0062 0.0091
    50 (0*50) 0.0019 0.0017 0.0021 0.0039 0.0085

     | Show Table
    DownLoad: CSV
    Table 4.  The CP values with confidence level 95%.
    n r Q η λ
    20 10 (3*1, 0*4, 7*1, 0*4) 0.897 0.970
    (10*1, 0*9) 0.976 0.987
    (2*5, 0*5) 0.937 0.974
    20 (0*20) 0.960 0.977
    30 10 (5*3, 0*2, 5*1, 0*4) 0.842 0.974
    (20*1, 0*9) 0.980 0.981
    (4*5, 0*5) 0.854 0.979
    20 (0*10, 2*5, 0*5) 0.826 0.958
    (10*1, 0*19) 0.977 0.978
    (1*10, 0*10) 0.930 0.977
    30 (0*30) 0.963 0.975
    50 15 (2*4, 1*6, 0*5) 0.889 0.974
    (35*1, 0*14) 0.976 0.982
    (7*5, 0*10) 0.832 0.986
    20 (5*6, 0*10, 0*4) 0.866 0.977
    (30*1, 0*19) 0.978 0.982
    (6*5, 0*15) 0.901 0.985
    30 (3*4, 0*7, 1*8, 0*11) 0.868 0.974
    (4*5, 0*25) 0.942 0.978
    (20*1, 0*29) 0.968 0.977
    50 (0*50) 0.957 0.977

     | Show Table
    DownLoad: CSV

    Balakrishnan and Sandhu [21] proposed an algorithm to produce progressive type-Ⅱ censored sample from any continuous distribution. The specific steps applied to the IWD are as follows:

    Step 1. Generate samples w1,w2,...,wr from U(0,1), where w1,w2,...,wr are independent.

    Step 2. Let vi=w(i+Qr+Qr1+...+Qri+1)1i and ui=1vrvr1...vri+1.

    Step 3. Let ti=F1(ui;ηreal,λreal), where F() is the CDF (1.2) of IWD, i=1,2,...,r.

    Then, t1<t2<...<tr are progressive type-Ⅱ censored data from IW(ηλ) with a censoring scheme Q. Calculation results can be found in Tables 13.

    MSE(ˆθ)=1NNi=1(ˆθiθreal)2. (5.1)

    From Table 1 to Table 3, we have the following conclusions:

    (i) Obviously, considering the same n and r, the censoring scheme Q has a great influence on MSE.

    (ii) Considering the same n, r and Q, the BE of η under SE loss function is the better than the MLE and IME.

    (iii) Considering the same n, r and Q, the BE λ under SE loss function is better than others. However, the BE of λ under SSE loss function is close to the BE under LINEX loss function.

    (iv) Considering the same n, r and Q, for the reliability R(t), MSEs of MLEs and BEs under SE and SSE loss functions are relatively close, while, MSEs of BEs under LINEX loss function are larger than others.

    From Table 4, We know that CP values for η and λ are all close to 95%.

    There is a set of real data from Dumonceaux and Antle [22], which represents the maximum flood level (in millions of cubic feet per second) of the Susquehanna River at Harrisburg, Pennsylvania over 20 four-year periods (1890–1969) as:

    0.265, 0.269, 0.297, 0.315, 0.324, 0.338, 0.379, 0.379, 0.392, 0.402, 0.412, 0.416, 0.418, 0.423, 0.449, 0.484, 0.494, 0.613, 0.654, 0.740.

    According to the data, we can plot the empirical CDF and the CDF of the IWD, as shown in Figure1. In the IWD, we using the BEs under SE loss function as the value of parameter, i.e. η=0.0336,λ=2.0431. According to Figure 1, we can see that the IWD can well model the data. Therefore, we can conclude that this distribution is valid.

    Figure 1.  The empirical CDF and the CDF of IWD.

    Now, the real data with censoring scheme (Q1,Q2,...,Q10)=(1,1,...,1) are as follows:

    0.265, 0.297, 0.324, 0.379, 0.392, 0.412, 0.418, 0.449, 0.494, 0.654.

    Before proceeding with the estimation, it is necessary to establish the existence and uniqueness of the maximum likelihood estimate. However, proving this can be a complicated process due to the nonlinearity of the system of Eq (2.5). For this reason, we visualize it through Figure 2, where L1 represents L(η,λ;t)η in Eq (2.3) and L2 represents L(η,λ;t)λ in Eq (2.4).

    Figure 2.  The Partial derivatives of log-likelihood function.

    From Figure 2, we know that the two curves intersect at only one point, indicating the presence of a unique MLE.

    The estimates and generalized confidence intervals that obtained by using these proposed methods are shown in Table 5.

    Table 5.  The results of real data analysis (t0=0.412).
    MLEs BEs IMEs GCIs
    SE SSE LINEX
    η 0.0582 0.0336 0.0196 0.0514 0.0238 (0.0101, 0.0732)
    λ 3.2009 2.0431 2.5710 2.9821 3.6513 (1.7307, 9.2871)
    R(t0) 0.6301 0.3859 0.3743 0.5330 0.4546

     | Show Table
    DownLoad: CSV

    In this paper, we have investigated the point estimation and interval estimation of parameters based on progressive type-Ⅱ censored sample for IWD. First, the Newton-Raphson iteration method is used to solve the likelihood equations of parameters and obtain their MLEs. Then, the BEs are derived based on SE and SSE loss functions, respectively. Finally, the IMEs are derived by generalized pivotal quantity. Additionally, the GCIs are also constructed by generalized pivotal quantity. Monte Carlo simulation is used to present the effect of the above estimators. The simulation results revealed that the estimators derived using Bayesian approach perform better than other methods in terms of MSE. Moreover, a real set of data is analyzed and the results coincide with simulation. Monte Carlo simulation results indicate that Bayesian estimation under the SE loss function works best among all the methods mentioned in this paper.

    The authors declare they have not used Artificial Intelligence (AI) tools in the creation of this article.

    This research was funded by National Natural Science Foundation of China (No. 71661012).

    The authors declare there is no conflict of interest.



    [1] A. Z. Keller, A. R. R. Kanath, Alternative reliability models for mechanical systems, In: Proceeding of the Third International Conference on Reliability and Maintainability, 1982,411–415.
    [2] C. Abhijit, C. Anindya, Use of the Fréchet distribution for UPV measurements in concrete, NDT E. Int., 52 (2012), 122–128. http://doi.org/10.1016/j.ndteint.2012.07.003 doi: 10.1016/j.ndteint.2012.07.003
    [3] C. Elio, F. P. De, N. L. P. Di, M. Fabio, Inverse loglogistic distribution for Extreme Wind Speed modeling: Genesis, identification and Bayes estimation, AIMS Energy, 6 (2018), 926–948. http://doi.org/10.3934/energy.2018.6.926 doi: 10.3934/energy.2018.6.926
    [4] A. O. Langlands, S. J. Pocock, G. R. Kerr, S. M. Gore, Long-term survival of patients with breast cancer: a study of the curability of the disease, Brit. Med. J., 2 (1979), 1247–1251. http://doi.org/10.1136/bmj.2.6200.1247 doi: 10.1136/bmj.2.6200.1247
    [5] Q. X. Bi, W. H. Gui, Bayesian and classical estimation of stress-strength reliability for inverse Weibull lifetime models, Algorithms, 10 (2017), 71–87. http://doi.org/10.3390/a10020071 doi: 10.3390/a10020071
    [6] W. S. A. El Azm, R. Aldallal, H. M. Aljohani, S. G. Nassr, Estimations of competing lifetime data from inverse Weibull distribution under adaptive progressively hybrid censored, Math. Biosci. Eng., 19 (2022), 6252–6276. http://doi.org/10.3934/mbe.2022292 doi: 10.3934/mbe.2022292
    [7] M. Alslman, A. Helu, Estimation of the stress-strength reliability for the inverse Weibull distribution under adaptive type-Ⅱ progressive hybrid censoring, PloS One, 17 (2022), e0277514. http://doi.org/10.1371/journal.pone.0277514 doi: 10.1371/journal.pone.0277514
    [8] A. I. Shawky, K. Khan, Reliability estimation in multicomponent stress-strength based on inverse Weibull distribution, Processes, 10 (2022), 364–401. http://doi.org/10.3390/pr10020226 doi: 10.3390/pr10020364
    [9] D. Kundu, H. Howlader, Bayesian inference and prediction of the inverse Weibull distribution for Type-Ⅱ censored data, Comput. Stat. Data. Anal., 54 (2010), 1547–1558. http://doi.org/10.1016/j.csda.2010.01.003 doi: 10.1016/j.csda.2010.01.003
    [10] F. G. Akgul, K. Yu, B. Senoglu, Classical and Bayesian inferences in step-stress partially accelerated life tests for inverse Weibull distribution under type-Ⅰ censoring, Strength Mater, 52 (2020), 480–496. http://doi.org/10.1007/s11223-020-00200-y doi: 10.1007/s11223-020-00200-y
    [11] A. Helu, H. Samawi, The inverse Weibull distribution as a failure model under various loss functions and based on progressive first-failure censored data, Qual. Technol. Quant. Manag., 12 (2016), 517–535. http://doi.org/10.1080/16843703.2015.11673434 doi: 10.1080/16843703.2015.11673434
    [12] K. Lee, Bayes and maximum likelihood estimation of uncertainty measure of the inverse Weibull distribution under generalized adaptive progressive hybrid censoring, Mathematics, 10 (2022), 4782–4802. http://doi.org/10.3390/math10244782 doi: 10.3390/math10244782
    [13] B. Xu, D. H. Wang, R. T. Wang, Estimator of scale parameter in a subclass of the exponential family under symmetric entropy loss, Northeast. Math. J., 24 (2008), 447–457. http://doi.org/10.3969/j.issn.1674-5647.2008.05.008 doi: 10.3969/j.issn.1674-5647.2008.05.008
    [14] L. X. Song, Y. S. Chen, J. M. Xu, Bayesian estimation of Poisson distribution parameter under scale squared error loss function, J. Lanzhou Univ. Tech., 34 (2008), 152–154. http://doi.org/10.3969/j.issn.1673-5196.2008.05.035 doi: 10.3969/j.issn.1673-5196.2008.05.035
    [15] H. R. Varian, A Bayesian approach to real estate assessment, ZELLNER A, FEINBERG S E., In Studies in Bayesian Econometrics and Statics In honor of L J. Savage, 1975,195–208.
    [16] D. V. Lindley, Approximate Bayesian methods, Trab. de Estad. y de Investig. Oper., 31 (1980), 223–245. http://doi.org/10.1007/bf02888353 doi: 10.1007/BF02888353
    [17] B. X. Wang, Statistical inference for Weibull distribution, Chinese J. Appl. Prob. Stat., 8 (1992), 357–364.
    [18] W. Luo, L. Z. Xue, J. W. Yao, X. F. Yu, Inverse moment methods for sufficient forecasting using high-dimensional predictors, Biometrika, 109 (2022), 473–487. http://doi.org/10.1093/biomet/asab037 doi: 10.1093/biomet/asab037
    [19] W. Qin, X. Yuan, An ensemble of inverse moment estimators for sufficient dimension reduction, Comput. Stat. Data. Anal., 161 (2021), 107241–107256. http://doi.org/10.1016/j.csda.2021.107241 doi: 10.1016/j.csda.2021.107241
    [20] S. Gao, J. Yu, W. H. Gui, Pivotal inference for the inverted exponentiated Rayleigh distribution based on progressive type-Ⅱ censored data, Am. J. Math. Manag. Sci., 39 (2020), 315–328. http://doi.org/10.1080/01966324.2020.1762142 doi: 10.1080/01966324.2020.1762142
    [21] N. Balakrishnan, R. A. Sandhu, A simple simulational algorithm for generating progressive Type-Ⅱ censored samples, Am. Stat., 49 (1995), 229–230. http://doi.org/10.2307/2684646 doi: 10.2307/2684646
    [22] R. Dumonceaux, C. E. Antle, Discrimination between the log-normal and the Weibull distributions, Technometrics, 15 (1973), 923–926. http://doi.org/10.2307/1267401 doi: 10.1080/00401706.1973.10489124
  • This article has been cited by:

    1. Mustafa M. Hasaballah, Yusra A. Tashkandy, Oluwafemi Samson Balogun, M. E. Bakr, Reliability analysis for two populations Nadarajah-Haghighi distribution under Joint progressive type-II censoring, 2024, 9, 2473-6988, 10333, 10.3934/math.2024505
    2. Mustafa M Hasaballah, Oluwafemi Samson Balogun, M E Bakr, Point and interval estimation based on joint progressive censoring data from two Rayleigh-Weibull distribution with applications, 2024, 99, 0031-8949, 085239, 10.1088/1402-4896/ad6107
  • Reader Comments
  • © 2023 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(1820) PDF downloads(113) Cited by(2)

Figures and Tables

Figures(2)  /  Tables(5)

Other Articles By Authors

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog