Research article

Estimation and prediction for two-parameter Pareto distribution based on progressively double Type-II hybrid censored data

  • Received: 18 February 2023 Revised: 10 April 2023 Accepted: 13 April 2023 Published: 26 April 2023
  • MSC : 62F10, 62F15

  • In this paper, a new censoring test plan called progressively double Type-II hybrid censoring scheme is introduced for the first time. Based on this type of censored data, the maximum likelihood estimates of the unknown parameters and reliability for the two-parameter Pareto distribution are obtained. Using the Bayesian method, the Bayesian estimates of the unknown parameters and reliability are obtained under the symmetric and asymmetric loss functions. The failure times of all withdrawn units are predicted using the classical and Bayesian methods, including the predictive values and the prediction intervals. The mean values and mean square errors of the estimators are calculated by Monte-Carlo simulation, and the mean square errors between them are compared, and the results show that all Bayesian estimates are better than the corresponding maximum likelihood estimates. Using a real data set, we compute the Bayesian estimates of the unknown parameters and reliability, and predict the observations of the censored units.

    Citation: Bing Long, Zaifu Jiang. Estimation and prediction for two-parameter Pareto distribution based on progressively double Type-II hybrid censored data[J]. AIMS Mathematics, 2023, 8(7): 15332-15351. doi: 10.3934/math.2023784

    Related Papers:

    [1] Ahmed R. El-Saeed, Ahmed T. Ramadan, Najwan Alsadat, Hanan Alohali, Ahlam H. Tolba . Analysis of progressive Type-Ⅱ censoring schemes for generalized power unit half-logistic geometric distribution. AIMS Mathematics, 2023, 8(12): 30846-30874. doi: 10.3934/math.20231577
    [2] Ahmed Elshahhat, Refah Alotaibi, Mazen Nassar . Statistical inference of the Birnbaum-Saunders model using adaptive progressively hybrid censored data and its applications. AIMS Mathematics, 2024, 9(5): 11092-11121. doi: 10.3934/math.2024544
    [3] Mohamed S. Eliwa, Essam A. Ahmed . Reliability analysis of constant partially accelerated life tests under progressive first failure type-II censored data from Lomax model: EM and MCMC algorithms. AIMS Mathematics, 2023, 8(1): 29-60. doi: 10.3934/math.2023002
    [4] 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
    [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] Hanan Haj Ahmad, Ehab M. Almetwally, Dina A. Ramadan . A comparative inference on reliability estimation for a multi-component stress-strength model under power Lomax distribution with applications. AIMS Mathematics, 2022, 7(10): 18050-18079. doi: 10.3934/math.2022994
    [7] Magdy Nagy, Khalaf S. Sultan, Mahmoud H. Abu-Moussa . Analysis of the generalized progressive hybrid censoring from Burr Type-Ⅻ lifetime model. AIMS Mathematics, 2021, 6(9): 9675-9704. doi: 10.3934/math.2021564
    [8] 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. AIMS Mathematics, 2024, 9(4): 10333-10352. doi: 10.3934/math.2024505
    [9] Mustafa M. Hasaballah, Oluwafemi Samson Balogun, M. E. Bakr . Frequentist and Bayesian approach for the generalized logistic lifetime model with applications to air-conditioning system failure times under joint progressive censoring data. AIMS Mathematics, 2024, 9(10): 29346-29369. doi: 10.3934/math.20241422
    [10] Amal Hassan, Sudhansu Maiti, Rana Mousa, Najwan Alsadat, Mahmoued Abu-Moussa . Analysis of competing risks model using the generalized progressive hybrid censored data from the generalized Lomax distribution. AIMS Mathematics, 2024, 9(12): 33756-33799. doi: 10.3934/math.20241611
  • In this paper, a new censoring test plan called progressively double Type-II hybrid censoring scheme is introduced for the first time. Based on this type of censored data, the maximum likelihood estimates of the unknown parameters and reliability for the two-parameter Pareto distribution are obtained. Using the Bayesian method, the Bayesian estimates of the unknown parameters and reliability are obtained under the symmetric and asymmetric loss functions. The failure times of all withdrawn units are predicted using the classical and Bayesian methods, including the predictive values and the prediction intervals. The mean values and mean square errors of the estimators are calculated by Monte-Carlo simulation, and the mean square errors between them are compared, and the results show that all Bayesian estimates are better than the corresponding maximum likelihood estimates. Using a real data set, we compute the Bayesian estimates of the unknown parameters and reliability, and predict the observations of the censored units.



    In the fields of reliability engineering and survival analysis, due to time or other constraints, it is often impossible to obtain the lifetimes of all tested units, so we can use a censored test scheme to obtain censored data. So far, many types of censoring schemes have been produced, and the corresponding statistical inference theories have been developed rapidly. Due to simplicity and applicability, Type-I and Type-II are the two most common and widely used censoring schemes. Type-I censoring is to stop the test when the preset time arrives, while Type-II censoring is to stop the test after a given number of failure data is obtained. A common feature of these two schemes is that the living units cannot be removed during the test. Sometimes it is necessary to withdraw some of the living units to observe their degradation, in this context, the progressive Type-II censoring scheme has been proposed. There are many research results based on the above three types of censored data. For more details, see references [1,2,3,4,5,6,7,8]. In some cases, for some high-reliability and long-life units, if the Type-II censoring scheme is adopted, the test time will be very long, and it will cost more money. In order to improve the test efficiency, another censoring scheme has been proposed, that is, the first-failure censoring scheme, which can save time and test costs. Based on the first-failure censored data, many scholars have studied the statistical inference on the parameters of various distributions, for example, Wu et al. [9,10] obtained maximum likelihood estimates and confidence intervals of the parameters for Gompertz and Burr XII distributions on the basis of the first-failure censored data. Wu and Kus [11] considered the advantages of the above censoring schemes, extended the first-failure censoring scheme, and proposed the progressive first-failure censoring scheme, and proved that this scheme had shorter expected test times than the progressive Type-II censoring scheme. Based on the characteristics of Type-I and Type-II censoring schemes, Epstein [12] first proposed a mixture of Type-I and Type-II censoring schemes, that is, Type-I hybrid censoring scheme. Later, Childs et al. [13] introduced another mixture of Type-I and Type-II censoring schemes, namely, Type-II hybrid censoring scheme. If n units are put into the life test at time zero, they have the following ordered lifetimes: X1:n,X2:n,...,Xn:n, and T represents the specified censoring time, r represents the number of failed units determined in advance. The Type-I hybrid censoring test is terminated at a random time T1=min(Xr:n,T), and the Type-II hybrid censoring test is terminated at a random time T2=max(Xr:n,T). The advantage of the Type-I hybrid censoring scheme is that the test time will not exceed T, and the disadvantage is that little failure data may be obtained. The advantage of the Type-II hybrid censoring scheme is that at least r failure data can be obtained, but the disadvantage is that the duration of the test cannot be controlled. For Type-I and Type-II hybrid censoring schemes, living units cannot be removed during the test. Therefore, Kundu et al. [14] extended the hybrid censoring scheme and proposed a progressive Type-I hybrid censoring scheme. At present, in the fields of reliability and survival analysis, hybrid censoring schemes are very popular in analyzing high-reliability life data. Interested readers may refer to references [15,16]. According to the existing censoring schemes, Long and Zhang [17] proposed a mixture of two Type-II censoring schemes, which are called double Type-II hybrid censoring scheme. When the lifetimes of the tested units follow two-parameter Pareto distribution, the statistical inference methods for the unknown parameters and reliability indicators are given based on this type of hybrid censored data. In this paper, we will generalize the double Type-II hybrid censoring scheme, propose a progressively double Type-II hybrid censoring scheme, and discuss the estimation of the unknown parameters and reliability under the two-parameter Pareto model, as well as the prediction of the failure times of evacuated units. The specific test scheme will be introduced in Section 2.

    One of the most important problems in life test is to predict future failure times based on observed data. In the early stages of the test, we can predict how expensive the test will be and whether measures need to be taken to adjust the test scheme. So far, many scholars have done a lot of work on prediction, for example, see references [18,19,20,21,22]. In this paper, we will further explore the prediction problems under the progressively double Type-II hybrid censored data.

    The rest of the paper is organized as follows: In Section 2, the two-parameter Pareto model and progressively double Type-II hybrid censoring scheme are introduced, and the maximum likelihood estimates of the unknown parameters and reliability are given. The Bayesian estimates of the unknown parameters and reliability are obtained in Section 3. In Section 4, the classical and Bayesian methods are used to predict failure times of the evacuated units. In Section 5, Monte-Carlo simulation is used to verify the goodness of the estimators. A real data set is analyzed in Section 6. We conclude the paper in Section 7.

    The two-parameter Pareto distribution was originally proposed as an income distribution, mainly used to analyze economic and natural phenomena, and later the distribution was also applied to the fields of reliability and survival analysis. Its cumulative distribution function and probability density function are respectively given as follow

    F(x;λ,θ)=1λθxθ,f(x;λ,θ)=θλθxθ+1,xλ, (2.1)

    where λ(>0) is the scale parameter and θ(>0) is the shape parameter. If X represents the lifetime of unit, the reliability function is

    R(x)=λθxθ. (2.2)

    In this paper, it is assumed that the lifetimes of the tested units follow the two-parameter Pareto distribution (2.1), and based on progressively double Type-II hybrid censored data, we will discuss the relevant estimation and prediction problems of this distribution.

    In the reliability test, progressively double Type-II hybrid censoring is a mixture of two progressive Type-II hybrid censoring schemes. The model is described as follows:

    Suppose that n independent and identically distributed units are put into the test, the time t0 and the positive integers m1,m2 are determined in advance, and m1<m2n are satisfied. When the first unit fails, the failure time is denoted as X1:n, and R1 non-failed units are removed from the remaining (n1) units. When the second unit fails, the failure time is denoted as X2:n, and R2 non-failed units are removed from the remaining (n2R1) units. By analogy, the failure time of the m1 -th unit is denoted as Xm1:n. If Xm1:nt0, the test is terminated at time Xm1:n, and all the Rm1=nm1m11i=1Ri units that have not failed are withdrawn from the test, where 0<X1:nX2:n...Xm1:n are the order times of failure. If Xm1:n<t0, the test is terminated when m2 units fail, and the Rm2=nm2m21i=1Ri units that have not failed are withdrawn from the test, where 0<X1:nX2:n...Xm2:n are the order times of failure. The number of removed units R1,R2,...,Rm2 can be determined in advance.

    Using the above test scheme, we can obtain the following two types of censored data:

    Case I: (X1:n,R1,X2:n,R2,...,Xm1:n,Rm1), if Xm1:nt0;

    Case II: (X1:n,R1,X2:n,R2,...,Xm2:n,Rm2), if Xm1:n<t0.

    Denote

    k={m1, Case Im2, Case II.

    Then the obtained progressively double Type-II hybrid censored data can be expressed as x_=(x1:n,R1,x2:n,R2,...,xk:n,Rk), if R1=R2=...=Rk1=0, Rk=nk, x_ is double Type-II hybrid censored data.

    Based on the experimental data x_, the likelihood function can be expressed as

    L(λ,θ)ki=1f(xi:n;λ,θ)[1F(xi:n;λ,θ)]RiI(x1:nλ)θkλnθexp{θ(ki=1lnxi:n+ki=1Rilnxi:n)}I(x1:nλ), (2.3)

    where I() is an indicator function.

    From (2.3), the maximum likelihood estimates of θ and λ can be obtained as

    ˜θ=kki=1lnxi:n+ki=1Rilnxi:nnlnx1:n,˜λ=x1:n.

    According to the invariance of the maximum likelihood estimation, the maximum likelihood estimate of the reliability function R(x) is

    ˜R(x)=(˜λ/x)˜θ.

    When λ is known, the maximum likelihood estimates of θ and R(x) are

    ˆθ=kki=1lnxi:n+ki=1Rilnxi:nnlnλ,ˆR(x)=(λ/x)ˆθ.

    In this section, when λ is known and θ is unknown, the prior distribution of θ is taken as Gamma distribution, and the Bayesian estimates of θ and reliability function are given. When both λ and θ are unknown, the Bayesian estimates of λ, θ and reliability function are obtained under three types of loss functions.

    Since λ is known, we only need to regard θ as a random variable, and the prior distribution of θ can be taken as Gamma distribution, and its probability density function is

    π(θ)=baΓ(a)θa1ebθ,θ>0, (3.1)

    where the hyper-parameters a>0,b>0, Γ() represents the Gamma function.

    According to reference [4], in order to ensure the robustness of Bayesian estimation, the value of a should satisfy 0<a<1, and the value of b should not be too large.

    From (2.3) and (3.1), according to the Bayesian formula, the posterior density function of θ can be obtained as

    π(θ|x_)=(A+b)k+aΓ(k+a)θk+a1eθ(A+b),θ>0, (3.2)

    where A=ki=1lnxi:n+ki=1Rilnxi:nnlnλ.

    In statistical decision theory and Bayesian analysis, the squared error loss function is a symmetric loss function that is often used, and an important advantage is that the Bayesian estimation of the estimated quantity can be easily calculated. The squared error loss is defined as

    L1(ϕ(β),δ)=[δϕ(β)]2,

    where δ is an estimate of ϕ(β).

    Under the squared error loss, the Bayesian estimate of ϕ(β) is

    ˆϕS=Eϕ[ϕ(β)|data], (3.3)

    where Eϕ() denotes posterior expectation with respect to the posterior density of ϕ(β).

    Using (3.2) and (3.3), the Bayesian estimates of θ and R(x) under the squared error loss are given, respectively, by

    ˆθBS=k+aA+b,
    ˆRBS(x)=(A+b)k+a(A+b+lnxlnλ)k+a.

    Based on the square error loss, the risks of overestimation and underestimation are the same. In some estimation and prediction problems, overestimation and underestimation will have different estimation risks, so the symmetric loss function may be unreasonable. Therefore, we consider two kinds of asymmetric loss functions, namely, the linear-exponential (LINEX) loss and general entropy loss functions, which are defined as

    L2(ϕ(β),δ)=ec(δϕ(β))c(δϕ(β))1,(c0),L3(ϕ(β),δ)(δϕ(β))qqln(δϕ(β))1.

    For the LINEX loss function, when c<0, the loss of underestimation is greater than that of overestimation, and when c>0, the opposite is true. For the general entropy loss function, when q<0, the loss of underestimation is greater than that of overestimation, and the opposite is true when q>0.

    Under the LINEX loss and general entropy loss, the Bayesian estimates of ϕ(β) are

    ˆϕL=1cln[Eϕ(ecϕ(β)|data)], (3.4)
    ˆϕG={Eϕ[(ϕ(β))q|data]}1/q (3.5)

    Using (3.2) and (3.4), when c>(A+b), the Bayesian estimates of θ and R(x) under the LINEX loss are given, respectively, by

    ˆθBL=k+acln(1+cA+b),
    ˆRBL(x)=1cln{(A+b)k+aj=0(c)jj![A+b+j(lnxlnλ)](k+a)}.

    Using (3.2) and (3.5), when q<k+a, the Bayesian estimates of θ and R(x) under the general entropy loss are given, respectively, by

    ˆθBG=[Γ(k+a)Γ(k+aq)]1/q(A+b)1,
    ˆRBG(x)=[1q(lnxlnλ)A+b]k+aq.

    In addition, the Bayesian credible interval for θ can be obtained from the posterior density (3.2).

    Since 2(A+b)θχ2(2k+2a), the 100(1γ)% equal-tail credible interval of θ is (L,U), where L=χ21γ/2(2k+2a)2(A+b), U=χ2γ/2(2k+2a)2(A+b), χ2γ(k) is the 100γ% right-tail percentile of the chi-squared distribution with k degrees of freedom.

    In most cases, both λ and θ are unknown. According to Bayesian theories, they are regarded as random variables, and prior distributions need to be given in advance. Here we take the prior distribution of λ as the noninformative prior distribution, namely

    π1(λ)=1λ,λ>0. (3.6)

    The prior distribution of θ is still taken as Gamma distribution, its probability density function is (3.1), and it is assumed that λ and θ are independent.

    Using the prior distributions (3.1) and (3.6), the joint posterior density of (λ,θ) is

    π(λ,θ|x_)=L(λ,θ)π1(λ)π(θ)+0x1:n0L(λ,θ)π1(λ)π(θ)dλdθ
    =n(B+bnlnx1:n)k+a1Γ(k+a1)θk+a1λnθ1eθ(B+b), (3.7)

    where B=ki=1lnxi:n+ki=1Rilnxi:n, 0<λx1:n,0<θ<+.

    Therefore, the posterior distribution of θ is Gamma distribution, and its probability density function is

    π(θ|x_)=x1:n0π(λ,θ|x_)dλ
    =(B+bnlnx1:n)k+a1Γ(k+a1)θk+a2eθ(B+bnlnx1:n),θ>0. (3.8)

    In addition, the posterior density of λ is

    π(λ|x_)=+0π(λ,θ|x_)dθ
    =n(k+a1)λ1(B+bnlnx1:n)k+a1(B+bnlnλ)(k+a),0<λx1:n. (3.9)

    So under the squared error loss, the Bayesian estimate of θ is

    ˜θBS=+0θπ(θ|x_)dθ=k+a1B+bnlnx1:n.

    Under the squared error loss, the Bayesian estimate of λ is

    ˜λBS=+0λπ(λ|x_)dλ
    =n(k+a1)(B+bnlnx1:n)k+a1x1:n0(B+bnlnλ)(k+a)dλ. (3.10)

    Under the squared error loss, the Bayesian estimate of R(x) is

    ˜RBS(x)=+0x1:n0λθxθπ(λ,θ|x_)dλdθ
    =nn+1(B+bnlnx1:n)k+a1Γ(k+a1)+0θk+a2exp{θ[B+b+lnx(n+1)lnx1:n]}dθ
    =nn+1[B+bnlnx1:nB+b+lnx(n+1)lnx1:n]k+a1.

    Using (3.4), (3.8) and (3.9), when c>(B+bnlnx1:n), the Bayesian estimates of θ and λ under the LINEX loss are given, respectively, by

    ˜θBL=k+a1cln(1+cB+bnlnx1:n),
    ˜λBL=1cln{n(k+a1)(B+bnlnx1:n)k+a1x1:n0λ1(B+bnlnλ)(k+a)ecλdλ}. (3.11)

    Because

    Eλ,θ[ecR(x)|x_]=Eλ,θ(ecλθxθ|x_)
    =n(B+bnlnx1:n)k+a1j=0(c)jj!(n+j)[B+b+jlnx(n+j)lnx1:n)](k+a1).

    The Bayesian estimate of R(x) under the LINEX loss is

    ˜RBL(x)=1cln{Eλ,θ[ecR(x)|x_]}
    =1cln{n(B+bnlnx1:n)k+a1j=0(c)jj!(n+j)[B+b+jlnx(n+j)lnx1:n)](k+a1)}.

    Using (3.5), (3.8) and (3.9), when q<k+a1, the Bayesian estimates of θ and λ under the general entropy loss are given, respectively, by

    ˜θBG=[Γ(k+a1)Γ(k+aq1)]1/q(B+bnlnx1:n)1,
    ˜λBG={n(k+a1)(B+bnlnx1:n)k+a1x1:n0λq1(B+bnlnλ)(k+a)dλ}1/q. (3.12)

    Because

    Eλ,θ[Rq(x)|x_]=Eλ,θ(λqθxqθ|x_)
    =n(B+bnlnx1:n)k+a1(nq)[B+bqlnx(nq)lnx1:n]k+a1,

    the Bayesian estimate of R(x) under the general entropy loss is

    ˜RBG(x)={n(B+bnlnx1:n)k+a1(nq)[B+bqlnx(nq)lnx1:n]k+a1}1/q.

    The approximate values of (3.10)–(3.12) can be obtained by numerical method, so as to obtain the Bayesian estimates of λ.

    According to the posterior density (3.9) of λ, its cumulative distribution function can be obtained as

    Fλ(y)=(B+bnlnx1:n)k+a1(B+bnlny)k+a1,0<yx1:n.

    Since Fλ(y)U(0,1), let u1,u2,...,uN be mutually independent random numbers from a uniform distribution U(0,1). Using the inverse transform method,

    yi=exp{1n[B+b(B+bnlnx1:n)u1/(k+a1)i]}, (i=1,2,...,N) are the random numbers from the density function (3.9). So we can obtain

    ˜λBS1NNi=1yi,˜λBL1cln(1NNi=1ecyi),˜λBG(1NNi=1yqi)1/q.

    Previously, we discussed the point estimation of the unknown parameters, and the Bayesian credible intervals for θ and λ will be given below.

    According to the posterior density (3.8) of θ, we can obtain

    2(B+bnlnx1:n)θχ2(2k+2a2),

    the 100(1γ)% equal-tail credible interval of θ is (θL,θU), where

    θL=χ21γ/2(2k+2a2)2(B+bnlnx1:n),θU=χ2γ/2(2k+2a2)2(B+bnlnx1:n).

    According to the posterior density (3.9), the 100(1γ)% equal-tail credible interval of λ is (λL,λU), where λL and λU should satisfy

    P(λ<λL)=λL0π(λ|x_)dλ=γ/2,P(λ>λU)=x1:nλUπ(λ|x_)dλ=γ/2. (3.13)

    According to (3.13), we can further obtain

    λL=exp{1n[B+b(B+bnlnx1:n)(γ/2)1k+a1]},
    λU=exp{1n[B+b(B+bnlnx1:n)(1γ/2)1k+a1]}.

    In this section, we will give the prediction methods of censored observations Zij, j=1,2,...,Ri, and i=1,2,...,k. For the given observation data x_, then the conditional density of Zij is

    f(zij|x_,λ,θ)=j(Rij)f(zij)[F(zij)F(xi:n)]j1[1F(zij)]Rij[1F(xi:n)]Ri, (4.1)

    where zij>xi:n, j=1,2,...,Ri.

    In this subsection, we consider the best unbiased predictor (BUP) to predict censored observation Zij. A statistic ˆzij is called a BUP of Zij, if the predictor error ˆzijZij has a mean zero and Var(ˆzijZij) is less than or equal to the variance of any unbiased predictor of Zij. Therefore, the BUP of Zij can be obtained as E(Zij|x_). In (4.1), if 1F(zij)1F(xi:n)=u, then u|x_Beta(Rij+1,j). If the BUP of Zij is denoted as ˆz(ij)BUP, we can obtain

    ˆz(ij)BUP=+xi:nzijf(zij|x_,λ,θ)dzij
    =xi:nBeta(Rij+1,j)10uRij1/θ(1u)j1du
    =xi:nBeta(Rij+1, j)Beta(Rij1/θ+1, j). (4.2)

    The above expression contains the unknown parameter θ, and the desired BUP can be obtained by substituting ˜θ into (4.2).

    In this subsection, we consider using the conditional median to predict censored observation Zij. According to the definition of the conditional median, for a given x_, the median of the conditional distribution of Zij is the conditional median predictor (CMP), denoted as ˆz(ij)CMP, it needs to satisfy

    P(Zijˆz(ij)CMP)=P(Zijˆz(ij)CMP).

    Using the conditional density function (4.1), it can be expressed as

    +ˆz(ij)CMPf(zij|x_,λ,θ)=0.5. (4.3)

    Substituting (2.1) into (4.1), the conditional density of Zij is

    f(zij|x_,λ,θ)=j(Rij)θj1m=0(1)m(j1m)x(Ri+mj+1)θi:nz(Ri+mj+1)θ1ij,zij>xi:n. (4.4)

    Then (4.3) can be transformed into

    j(Rij)j1m=0(1)m(j1m)1Ri+mj+1(xi:nˆz(ij)CMP)θ(Ri+mj+1)=0.5. (4.5)

    Therefore, the CMP ˆz(ij)CMP of Zij is the solution of Eq (4.5), and θ is replaced by its maximum likelihood estimate ˜θ.

    In the previous subsections we used the classical methods to predict Zij. In this subsection we consider the Bayesian method to predict Zij. Using the posterior density (3.7) of (λ,θ), the corresponding posterior prediction density is

    h(zij|x_)=+0x1:n0f(zij|x_,λ,θ)π(λ,θ|x_)dλdθ
    =(k+a1)(B+bnlnx1:n)k+a1j(Rij)j1m=0(1)m(j1m)
    ×z1ij[B+bnlnx1:n+(Ri+mj+1)(lnzijlnxi:n)](k+a). (4.6)

    According to (4.6), the Bayesian posterior survival function is

    S(t|x_)=+th(zij|x_)dzij
    =(B+bnlnx1:n)k+a1j(Rij)j1m=0(1)m(j1m)
    ×1Ri+mj+1[B+bnlnx1:n+(Ri+mj+1)(lntlnxi:n)](k+a)+1,t>xi:n. (4.7)

    According to the definition of the median, if the Bayesian median predictor (BMP) of Zij is ˆz(ij)BMP, it needs to satisfy

    S(ˆz(ij)BMP|x_)=0.5,

    that is

    (B+bnlnx1:n)k+a1j(Rij)j1m=0(1)m(j1m)
    ×1Ri+mj+1[B+bnlnx1:n+(Ri+mj+1)(lnˆz(ij)BMPlnxi:n)](k+a)+1=0.5.

    In the previous subsections, we obtained point prediction of Zij using a variety of methods. Prediction intervals of censored observations will be constructed below, and two types of prediction intervals are obtained in this subsection using the classical method and the Bayesian method, respectively.

    According to the conditional density (4.4), the predictive survival function can be obtained as

    S(t|x_,λ,θ)=+tf(zij|x_,λ,θ)dzij
    =j(Rij)j1m=0(1)m(j1m)1Ri+mj+1(xi:nt)(Ri+mj+1)θ,t>xi:n.

    Therefore, the 100(1γ)% classical prediction interval of Zij is (ˆz(ij)L,ˆz(ij)U), and the lower bound ˆz(ij)L and the upper bound ˆz(ij)U should satisfy

    j(Rij)j1m=0(1)m(j1m)1Ri+mj+1(xi:nˆz(ij)L)(Ri+mj+1)θ=1γ/2,
    j(Rij)j1m=0(1)m(j1m)1Ri+mj+1(xi:nˆz(ij)U)(Ri+mj+1)θ=γ/2,

    where θ is replaced by its maximum likelihood estimate ˜θ.

    According to the Bayesian posterior survival function (4.7), the 100(1γ)% Bayesian prediction interval of Zij is (˜z(ij)L,˜z(ij)U), and ˜z(ij)L and ˜z(ij)U should satisfy

    (B+bnlnx1:n)k+a1j(Rij)j1m=0(1)m(j1m)
    ×1Ri+mj+1[B+bnlnx1:n+(Ri+mj+1)(ln˜z(ij)Llnxi:n)](k+a)+1=1γ/2,
    (B+bnlnx1:n)k+a1j(Rij)j1m=0(1)m(j1m)
    ×1Ri+mj+1[B+bnlnx1:n+(Ri+mj+1)(ln˜z(ij)Ulnxi:n)](k+a)+1=γ/2.

    In this section, we will use Monte-Carlo simulation to verify the properties of the estimators. The simulated samples come from the progressively double Type-II hybrid censoring scheme of the two-parameter Pareto distribution, and λ=6,θ=2. When the values of m1, m2, (R1,R2,...,Rk) and t0 are given under small sample size (n=20), medium sample size (n=40), and large sample size (n=60), respectively, the progressively double Type-II hybrid censored data with three sample sizes can be obtained. The values of the hyper-parameters are taken as (a,b)=(1,1), and c=1,q=1 in the loss functions. The values of the estimators are calculated based on the censored data. Each estimator is simulated 10000 times to calculate mean value (MV) and mean square error (MSE), and the formula of MSE is

    MSE(ˆϕ)=110000(ˆϕϕ)2,

    where ˆϕ is an estimate of ϕ.

    For convenience, short notation is used to represent different, for example, scheme (3,0,0,0,0) is denoted as (3,04). The MVs and MSEs for all point estimates of λ,θ and R(x) are listed in Tables 15. From the values in the Tables, it is easily observed that when n is fixed, the MSEs of all estimators decrease as t0 increases. Furthermore, when t0 is fixed, the MSEs of all estimators decrease as n increases. In terms of MSEs, all Bayesian estimates are better than the corresponding maximum likelihood estimates under the same condition, and the differences between them decrease rapidly as n increases. Especially in the case of small sample size, the advantage of Bayesian estimation is more obvious. In general, for three types of Bayesian estimates of θ and R(x), the MSE is minimal under the LINEX loss. For the Bayesian estimation of λ, the MSE is minimal under the general entropy loss. Under the same condition, when λ is unknown, the MSEs of the point estimates of θ and R(x) is larger than the MSEs when λ is known. From the MVs of the point estimates, the maximum likelihood estimates of the unknown parameters θ and λ are greater than their Bayesian estimates. The maximum likelihood estimate of the reliability function is less than three Bayesian estimates.

    Table 1.  MVs and MSEs of point estimates for θ when λ is known.
    n m1 m2 (R1,R2,...,Rk) t0 ˆθ ˆθBS ˆθBL ˆθBG
    20 10 14 (0*9, 10) or (0*9, 3, 0*3, 3) 8 MV 2.13812 1.91591 1.75973 1.74171
    MSE 0.32570 0.20271 0.19925 0.22839
    12 MV 1.88524 1.77095 1.67082 1.65297
    MSE 0.28010 0.19222 0.18354 0.22487
    40 20 28 (0*19, 20) or (0*19, 6, 0*7, 6) 8 MV 2.02012 1.91813 1.83252 1.82686
    MSE 0.21293 0.16274 0.15797 0.17158
    12 MV 2.03174 1.95786 1.89290 1.89014
    MSE 0.13764 0.11101 0.10686 1.87791
    60 30 42 (0*29, 30) or (0*29, 8, 0*11, 10) 8 MV 2.00764 1.94051 1.88055 1.87791
    MSE 0.14173 0.11815 0.11490 0.12226
    12 MV 2.03286 1.98336 1.93827 1.93738
    MSE 0.08793 0.07465 0.07148 0.07494

     | Show Table
    DownLoad: CSV
    Table 2.  MVs and MSEs of point estimates for R(x) when λ is known (x=7.5).
    n m1 m2 (R1,R2,...,Rk) t0 ˆR ˆRBS ˆRBL ˆRBG
    20 10 14 (0*9, 10) or (0*9, 3, 0*3, 3) 8 MV 0.62604 0.66065 0.65742 0.64968
    MSE 0.00689 0.00432 0.00428 0.00448
    12 MV 0.65946 0.67904 0.67543 0.67195
    MSE 0.00454 0.00406 0.00365 0.00379
    40 20 28 (0*19, 20) or (0*19, 6, 0*7, 6) 8 MV 0.64046 0.65712 0.65401 0.65138
    MSE 0.00409 0.00337 0.00336 0.00342
    12 MV 0.63761 0.64990 0.64625 0.64562
    MSE 0.00265 0.00225 0.00224 0.00229
    60 30 42 (0*29, 30) or (0*29, 8, 0*11, 10) 8 MV 0.64111 0.65233 0.64901 0.64838
    MSE 0.00272 0.00239 0.00237 0.00241
    12 MV 0.63667 0.64501 0.64300 0.64207
    MSE 0.00166 0.00146 0.00146 0.00148

     | Show Table
    DownLoad: CSV
    Table 3.  MVs and MSEs of point estimates for θ when λ is unknown.
    n m1 m2 (R1,R2,...,Rk) t0 ˜θ ˜θBS ˜θBL ˜θBG
    20 10 14 (0*9, 10) or (0*9, 3, 0*3, 3) 8 MV 2.23162 1.90169 1.73313 1.71152
    MSE 0.37844 0.23303 0.22707 0.26414
    12 MV 2.00654 1.74541 1.64163 1.62074
    MSE 0.32407 0.18069 0.21918 0.24375
    40 20 28 (0*19, 20) or (0*19, 6, 0*7, 6) 8 MV 2.11199 1.90141 1.81309 1.80634
    MSE 0.25496 0.16798 0.16572 0.18034
    12 MV 2.09806 1.94752 1.88105 1.87796
    MSE 0.15981 0.11365 0.11056 0.11801
    60 30 42 (0*29, 30) or (0*29, 8, 0*11, 10) 8 MV 2.06281 1.92611 1.86518 1.86191
    MSE 0.15110 0.11631 0.11539 0.12265
    12 MV 2.07338 1.97403 1.92826 1.92703
    MSE 0.09413 0.07295 0.07074 0.07421

     | Show Table
    DownLoad: CSV
    Table 4.  MVs and MSEs of point estimates for R(x) when λ is unknown (x=7.5).
    n m1 m2 (R1,R2,...,Rk) t0 ˜R ˜RBS ˜RBL ˜RBG
    20 10 14 (0*9, 10) or (0*9, 3, 0*3, 3) 8 MV 0.62867 0.66028 0.65698 0.64910
    MSE 0.00871 0.00441 0.00434 0.00457
    12 MV 0.67303 0.67696 0.67346 0.66931
    MSE 0.00597 0.00431 0.00394 0.00404
    40 20 28 (0*19, 20) or (0*19, 6, 0*7, 6) 8 MV 0.64217 0.65660 0.65311 0.65082
    MSE 0.00451 0.00332 0.00314 0.00337
    12 MV 0.64290 0.64897 0.64582 0.64460
    MSE 0.00288 0.00223 0.00213 0.00227
    60 30 42 (0*29, 30) or (0*29, 8, 0*11, 10) 8 MV 0.64226 0.65203 0.65101 0.64807
    MSE 0.00286 0.00235 0.00221 0.00237
    12 MV 0.63998 0.64431 0.64232 0.64133
    MSE 0.00175 0.00145 0.00138 0.00148

     | Show Table
    DownLoad: CSV
    Table 5.  MVs and MSEs of point estimates for λ.
    n m1 m2 (R1,R2,...,Rk) t0 ˜λ ˜λBS ˜λBL ˜λBG
    20 10 14 (0*9, 10) or (0*9, 3, 0*3, 3) 8 MV 6.14564 5.95291 5.93523 5.95753
    MSE 0.04373 0.02135 0.02324 0.02056
    12 MV 6.14564 5.96641 5.94884 5.96968
    MSE 0.04373 0.01873 0.02016 0.01750
    40 20 28 (0*19, 20) or (0*19, 6, 0*7, 6) 8 MV 6.06610 5.96132 5.95652 5.96550
    MSE 0.00822 0.00765 0.00824 0.00758
    12 MV 6.06610 5.97758 5.97013 5.97908
    MSE 0.00822 0.00714 0.00789 0.00702
    60 30 42 (0*29, 30) or (0*29, 8, 0*11, 10) 8 MV 6.04169 5.97864 5.96840 5.98213
    MSE 0.00307 0.00302 0.00324 0.00286
    12 MV 6.04169 5.98151 5.97204 5.98540
    MSE 0.00307 0.00294 0.00311 0.00279

     | Show Table
    DownLoad: CSV

    Next, we will analyze the example from reference [22], in which the failure data follow Type-II Pareto distribution. After transformation, the failure data from the two-parameter Pareto distribution can be obtained, which are arranged in ascending order as follows: 0.5009, 0.5040, 0.5142, 0.5221, 0.5261, 0.5418, 0.5473, 0.5834, 0.6091, 0.6252, 0.6404, 0.6498, 0.6750, 0.7031, 0.7099, 0.7168, 0.7918, 0.8465, 0.9035, 1.1143.

    If we take m1=10, m2=14, t0=0.7, and the number of units progressively removed is (R1,R2,...,Rk)=(09,3,03,3), then the censored data are obtained as 0.5009, 0.5040, 0.5142, 0.5221, 0.5261, 0.5418, 0.5473, 0.5834, 0.6091, 0.6252, 0.6404, 0.6750, 0.7031, 0.7168.

    Using the conclusions in this paper, when a,b,c and q are taken different values, we can calculate the Bayesian estimates of unknown parameters and reliability under three types of loss functions, and the obtained estimates are shown in Tables 6 and 7. According to the test data, the failure observations of the removed units are predicted, including point prediction and interval prediction, and the computational results are shown in Table 8.

    Table 6.  Bayesian estimates of θ and λ.
    (a,b) ˜θBS ˜λBS (c,q) ˜θBL ˜λBL ˜θBG ˜λBG
    (1, 1) 2.89176 0.49175 (1, 2) 2.62876 0.49172 2.78657 0.49145
    (2, -1) 2.42054 0.49167 3.09831 0.49175
    (-1, -2) 3.23918 0.49179 3.19992 0.49185
    (-2, 1) 3.73041 0.49185 2.89176 0.49156
    (1, 2) 2.39671 0.48991 (1, 2) 2.21233 0.48985 2.30953 0.48947
    (2, -1) 2.06115 0.48978 2.56790 0.48991
    (-1, -2) 2.62876 0.48997 2.65212 0.49005
    (-2, 1) 2.93397 0.49004 2.39671 0.48962

     | Show Table
    DownLoad: CSV
    Table 7.  Bayesian estimates of R(x) (x=0.6).
    (a,b) ˜RBS (c,q) ˜RBL ˜RBG
    (1, 1) 0.57045 (1, 2) 0.56709 0.55145
    (2, -1) 0.56372 0.57045
    (-1, -2) 0.57379 0.57629
    (-2, 1) 0.57712 0.55805
    (1, 2) 0.62194 (1, 2) 0.61904 0.60696
    (2, -1) 0.61611 0.62194
    (-1, -2) 0.62483 0.62658
    (-2, 1) 0.62768 0.61215

     | Show Table
    DownLoad: CSV
    Table 8.  Prediction of Zij (a=1,b=1,γ=0.05).
    i j ˆz(ij)BUP ˆz(ij)CMP ˆz(ij)BMP (ˆz(ij)L,ˆz(ij)U) (˜z(ij)L,˜z(ij)U)
    10 1 0.68813 0.66612 0.67856 (0.62665, 0.87605) (0.62702, 1.01671)
    2 0.79755 0.75616 0.79946 (0.64239, 1.19508) (0.64642, 1.63324)
    3 1.09914 0.96407 1.09542 (0.68745, 2.31989) (0.70087, 4.43453)
    14 1 0.78859 0.76372 0.77801 (0.71846, 1.00445) (0.71889, 1.16586)
    2 0.91441 0.86695 0.91650 (0.73649, 1.37017) (0.74109, 1.87320)
    3 1.26018 1.10534 1.25543 (0.78815, 2.66005) (0.80359, 5.08850)

     | Show Table
    DownLoad: CSV

    It can be seen from Tables 6 and 7 that under the LINEX loss, the Bayesian estimates of θ and R(x) are larger when c is negative than when c is positive. Also, under the general entropy loss, the Bayesian estimates of θ and R(x) are larger when q is negative than when q is positive. For different values of a,b,c and q, the Bayesian estimates of λ are very close under three types of loss functions. In Table 8, comparing the point prediction under three types of loss functions, it is found that the CMP is the smallest. The length of the Bayesian prediction intervals of Zij is greater than that of the classical prediction interval, and the real observations are within the prediction intervals.

    In this paper, based on the progressively double Type-II hybrid censored data, the statistical inference of the two-parameter Pareto distribution is studied by using the classical and Bayesian methods. The maximum likelihood estimates of the unknown parameter(s) and reliability are obtained when the scale parameter is known and unknown, respectively. In the Bayesian method, we obtain the Bayesian estimates of the unknown parameter(s) and reliability under the squared loss, LINEX loss and general entropy loss, respectively. Since the Bayesian estimation of λ cannot be obtained in an explicit form, a Monte-Carlo simulation method is proposed to obtain its Bayesian estimation, and we also obtain the Bayesian credible intervals of the unknown parameters. The point prediction and interval prediction of the failure observations of the withdrawn units are carried out by the classical and Bayesian methods, and the point prediction includes the best unbiased predictor, the conditional median predictor and the Bayesian median predictor. Simulation results show that, on the basis of MSE, the Bayesian estimation is better than the corresponding maximum likelihood estimation. Based on a real data set, we calculate the Bayesian estimates of the unknown parameters and reliability, and predict the observations of the censored units. We mainly consider the application of the progressively double Type-II hybrid censoring scheme in the two-parameter Pareto model, and this scheme can also be applied to other life distributions.

    The authors would like to thank the Editor and the reviewers for constructive comments and helpful suggestions that have led to a substantial improvement to an earlier version of the paper. The authors are so thankful for the support from the special fund for Nature Science Foundation of Hubei province, China (No. 2022CFB527) and Campus level scientific research platform of Jingchu University of Technology: Data Analysis Science Laboratory.

    All authors declare no conflicts of interest.



    [1] X. J. Zhu, N. Balakrishnan, C. Z. Feng, J. C. Ni, N. X. Yu, W. L. Zhou, Exact likelihood-ratio tests for joint type-II censored exponential data, Statistics, 54 (2020), 636–648. https://doi.org/10.1080/02331888.2020.1764559 doi: 10.1080/02331888.2020.1764559
    [2] H. Nagatsuka, N. Balakrishnan, A consistent method of estimation for the three-parameter lognormal distribution based on Type-II right censored data, Commun. Stat. Theor. M., 45 (2016), 5693–5708. http://doi.org/10.1080/03610926.2014.948205 doi: 10.1080/03610926.2014.948205
    [3] I. Basak, N. Balakrishnan, A note on the prediction of censored exponential lifetimes in a simple step-stress model with Type-II censoring, Calcutta Statistical Association Bulletin, 70 (2018), 57–73. https://doi.org/10.1177/0008068318769506 doi: 10.1177/0008068318769506
    [4] M. Han, E-Bayesian estimation and its E-posterior risk of the exponential distribution parameter based on complete and type I censored samples, Commun. Stat. Theor. M., 49 (2020), 1858–1872. https://doi.org/10.1080/03610926.2019.1565837 doi: 10.1080/03610926.2019.1565837
    [5] J. R. Ren, W. H. Gui, A statistical inference for generalized Rayleigh model under Type-II progressive censoring with binomial removals, J. Syst. Eng. Electron., 31 (2020), 206–223. https://doi.org/10.21629/JSEE.2020.01.20 doi: 10.21629/JSEE.2020.01.20
    [6] I. Basak, N. Balakrishnan, Prediction of censored exponential lifetimes in a simple step-stress model under progressive Type II censoring, Comput. Stat., 32 (2017), 1665–1687. https://doi.org/10.1007/s00180-016-0684-0 doi: 10.1007/s00180-016-0684-0
    [7] A. A. Soliman, A. H. A. Ellah, N. A. Abou-Elheggag, R. M. El-Sagheer, Bayesian and frequentist prediction using progressive Type-II censored with binomial removals, Intelligent Information Management, 5(2013), 162–170. http://doi.org/10.4236/iim.2013.55017
    [8] M. Rezapour, M. H. Alamatsaz, N. Balakrishnan, Distribution of the number of observations greater than the ith dependent progressively Type-II censored order statistic and its use in goodness-of-fit testing, Commun. Stat. Theor. M., 44 (2015), 2517–2529. http://doi.org/10.1080/03610926.2015.1043796 doi: 10.1080/03610926.2015.1043796
    [9] J. W. Wu, W. L. Hung, C. H. Tsai, Estimation of the parameters of the Gompertz distribution under the first failure-censored sampling plan, Statistics, 37 (2003), 517–525. http://doi.org/10.1080/02331880310001598864 doi: 10.1080/02331880310001598864
    [10] J. W. Wu, H. Y. Yu, Statistical inference about the shape parameter of the Burr type XII distribution under the failure-censored sampling plan, Appl. Math. Comput., 163 (2005), 443–482. https://doi.org/10.1016/j.amc.2004.02.019 doi: 10.1016/j.amc.2004.02.019
    [11] S. J. Wu, C. Kus, On estimation based on progressive first-failure-censored sampling, Comput. Stat. Data An., 53 (2009), 3659–3670. https://doi.org/10.1016/j.csda.2009.03.010 doi: 10.1016/j.csda.2009.03.010
    [12] B. Epstein, Truncated life tests in the exponential case, Ann. Math. Statist., 25 (1954), 555–564.
    [13] A. Childs, B. Chandrasekar, N. Balakrishnan, D. Kundu, Exact likelihood inference based on Type-I and Type-II hybrid censored samples from the exponential distribution, Ann. Inst. Stat. Math., 55 (2003), 319–330. https://doi.org/10.1007/BF02530502 doi: 10.1007/BF02530502
    [14] D. Kundu, A. Joarder, Analysis of Type-II progressively hybrid censored data, Comput. Stat. Data An., 50 (2006), 2509–2528. https://doi.org/10.1016/j.csda.2005.05.002 doi: 10.1016/j.csda.2005.05.002
    [15] C. T. Lin, Y. Y. Hsu, S. Y. Lee, N. Balakrishnan, Inference on constant stress accelerated life tests for log-location-scale lifetime distributions with type-I hybrid censoring, J. Stat. Comput. Sim., 89 (2019), 720–749. https://doi.org/10.1080/00949655.2019.1571591 doi: 10.1080/00949655.2019.1571591
    [16] N. Balakrishnan, A. R. Shafay, One-and two-sample Bayesian prediction intervals based on Type-II hybrid censored data, Commun. Stat. Theor. M., 41 (2012), 1511–1531. http://doi.org/10.1080/03610926.2010.543300 doi: 10.1080/03610926.2010.543300
    [17] B. Long, Z. Z. Zhang, Statistical analysis of two-parameter Pareto distribution under double Type-II hybrid censoring scheme, Acta Math. Sci., 42 (2022), 269–281. https://doi.org/10.3969/j.issn.1003-3998.2022.01.021 doi: 10.3969/j.issn.1003-3998.2022.01.021
    [18] B. Pradhan, D. Kundu, Bayes estimation and prediction of the two-parameter gamma distribution, J. Stat. Comput. Sim., 81 (2011), 1187–1198. http://doi.org/10.1080/00949651003796335 doi: 10.1080/00949651003796335
    [19] R. Valiollahi, A. Asgharzadeh, M. Z. Raqab, Prediction of future failures times based on Type-I hybrid censored samples of random sample sizes, Commun. Stat. Simul. C., 48 (2019), 109–125. http://doi.org/10.1080/03610918.2017.1375519 doi: 10.1080/03610918.2017.1375519
    [20] R. Valiollahi, A. Asgharzadeh, D. Kundu, Prediction of future failures for generalized exponential distribution under Type-I or Type-II hybrid censoring, Braz. J. Probab. Stat., 31 (2017), 41–61. https://www.jstor.org/stable/24913861
    [21] A. Asgharzadeh, R. Valiollahi, D. Kundu, Prediction for future failures in Weibull distribution under hybrid censoring, J. Stat. Comput. Sim., 85 (2015), 824–838. http://doi.org/10.1080/00949655.2013.848451 doi: 10.1080/00949655.2013.848451
    [22] A. M. Nigm, E. K. Al-Hussaini, Z. F. Jaheen, Bayesian one-sample prediction of future observations under Pareto distribution, Statistics, 37 (2003), 527–536. https://doi.org/10.1080/02331880310001598837 doi: 10.1080/02331880310001598837
  • 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
    3. Weisheng Wang, Wenhui Shi, Dongliang Nan, Yinzhang Peng, Qinghua Wang, Yankai Zhu, New energy output prediction and demand response optimization based on LSTM-BN, 2025, 2252-4940, 10.61435/ijred.2025.60632
  • 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(1554) PDF downloads(94) Cited by(3)

Figures and Tables

Tables(8)

Other Articles By Authors

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog