Abstract
Gamma stochastic process has been proposed to replace Brownian motion and geometric Brownian motion to characterize the degradation measurements from an accelerated degradation test (ADT). Tsai et al. [27] applied bivariate Gamma process to model the two-variable ADT under the independent assumption. In this chapter, we consider three bivariate Gamma processes utilizing the Clayton, Frank, and Gumbel copulas to describe the dependence characteristics of bivariate Gamma variables. Owing to the complex structure of modeling, the differentiation-based likelihood estimation of the copulas are not always tractable. Bayesian analysis using Markov chain Monte Carlo method is an effective alternative to implement parameter estimation and model comparisons. Extensive simulation studies that calculate the mean square errors (MSEs) of the derived estimates are conduced to show the efficiency of the proposed method. Three data sets from the Clayton, Frank, and Gumbel copulas are analyzed and used for the demonstration of model selections via Bayesian approach.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Similar content being viewed by others
References
Boulanger, M., Escobar, L. A.: Experimental design for a class of accelerated degradation tests. Technometrics 36, 260–272 (1994)
Chen, M. H.: Computing marginal likelihoods from a single MCMC output. Statistica Neerlandica 59, 16–29 (2005)
Chiang, J.-Y, Zhu, J., Lin, Y.-J., Lio, Y.L., Tsai, T.-R.: Inference from two-variable degradation data using genetic algorithm and Markov Chain Monte Carlo Methods. Int. J. Inform. Manag. Sci. 29, 235–256 (2018)
Chib, S., Greenberg, E.: Understanding the metropolis-hastings algorithm. Am. Stat. 49, 327–335 (1995)
Cowles, M.K., Carlin, B.P.: Markov chain Monte Carlo convergence diagnostics: a comparative review. J. Am. Stat. Assoc. 91(434), 883–904 (1996)
de Valpine, P., Paciorek, C., Turek, D., Michaud, N., Anderson-Bergman, C., Obermeyer, F., Wehrhahn Cortes, C., RodrÃguez, A., Temple Lang, D., Paganin, S.: NIMBLE: MCMC, Particle Filtering, and Programmable Hierarchical Modeling (2021). https://cran.r-project.org/package=nimble
de Valpine, P., Paciorek, C., Turek, D., Michaud, N., Anderson-Bergman, C., Obermeyer, F., Wehrhahn Cortes, C., RodrÃguez, A., Temple Lang, D., Paganin, S.: NIMBLE User Manual (2021). r-nimble.org
Gelman, A., Carlin, J.B., Stern, H.S., Dunson, D., Vehtari, A., Rubin, D.B.: Bayesian Data Analysis (3rd edn.) Chapman & Hall, London (2013)
Geman, S., Geman, D.: Stochastic relaxation, gibbs distributions, and the bayesian restoration of images. IEEE Trans. Pattern Anal. Math. Intell. 6, 721–741 (1984)
Gertsbackh, I.B., Kordonskiy, K.B.: Models of Failure. Springer, New York (1969)
Guan, Q., Tang, Y.-C.: Optimal design of accelerated degradation test based on gamma process models. Chinese J. Appl. Probab. Stat. 29, 213–224 (2013)
Hastings, W.K.: Monte Carlo sampling methods using Markov Chains and their applications. Biometrika 57, 97–109 (1970)
Liao, C.-M., Tseng, S.-T.: Optimal design for step-stress accelerated degradation tests. IEEE Trans. Reliab. 55, 59–66 (2006)
Lim, H., Yum, B.-J.: Optimal design of accelerated degradation tests based on Wiener process models. J. Appl. Stat. 38, 309–325 (2011)
Ling, M.H., Tsui, K.L., Balakrishnan, N.: Accelerated degradation analysis for the quality of a system based on the gamma process. IEEE Trans. Reliab. 64, 463–472 (2015)
Lu, C.J., Meeker, W.Q.: Using degradation measures to estimate a time-to-failure distribution. Technometrics 35, 161–174 (1993)
Metropolis, N., Rosenbluth, A.W., Rosenbluth, M.N., Teller, A.H., Teller, E.: Equations of state calculations by fast computing machines. J. Chem. Phys. 21, 1087–1092 (1953)
Padgett, W.J., Tomlinson, M.A.: Inference from accelerated degradation and failure data based on Gaussian process models. Lifetime Data Anal. 10, 191–206 (2004)
Park, C., Padgett, W.J.: Accelerated degradation models for failure based on geometric Brownian motion and gamma process. Lifetime Data Anal. 11, 511–527 (2005)
Park, C., Padgett, W.J.: Stochastic degradation models with several accelerating variables. IEEE Trans. Reliab. 55, 379–390 (2006)
Park, C., Padgett, W.J.: Cumulative damage models for failure with several accelerating variables. Quality Technol. Quant. Manag. 4, 17–34 (2007)
Peng, J.-Y.: A note on optimal allocations for the second elementary symmetric function with applications for optimal reliability design. Nav. Res. Logist. 59, 278–284 (2012)
Tan, M.T., Tian, G.L., Ng, K.W.: Bayesian Missing Data Problems: EM, Data Augmentation and Noniterative Computation. Chapman & HalI/CRC CRC Press, Florida (2010)
Tsai, T.-R., Lin, C.-W., Sung, Y.-L., Chou, P.-T., Chen C.-L., Lio, Y.L.: Inference from lumen degradation data under Wiener diffusion process. IEEE Trans. Reliab. 61, 710–718 (2012)
Tsai, C.-C., Tseng, S.-T., Balakrishnan, N.: Optimal burn-in policy for highly reliable products using gamma degradation process. IEEE Trans. Reliab. 60, 338–350 (2013)
Tsai, T.-R., Lio, Y. L., Jiang, N.: Optimal decision on the accelerated degradation test plan under the Wiener process. Qual. Technol. Quantit. Manag. 11, 461–470 (2014)
Tsai, T.-R., Sung, W-Y, Lio, Y. L., Chang, S., Lu, J.-C.: Optimal two-variable accelerated degradation test plan for gamma degradation processes. IEEE Trans. Reliab. 65, 459–468 (2016)
Tseng, S.-T., Balakrishnan N., Tsai, C.-C.: Optimal step-stress accelerated degradation test plan for gamma degradation process. IEEE Trans. Reliab. 58, 234–245 (2011)
Whitmore, G.A.: Estimating degradation by a Wiener diffusion process subject to measurement error. Lifetime Data Anal. 1, 307–319 (1995)
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Appendix
Appendix
Below are the R codes via NIMBLE package in Bayesian using Gibbs sampling (BUGS) language to estimate the parameters (s1, r2, s2, r2), including copula parameter (theta).
# install.packages("copula")         ## R copula package                                         (r-nimble.org) # install.packages("nimble", type="source") #    NIMBLE needs a C++ compiler and the standard utility 'make'      to generate and compile C++ library(nimble)                  ## call NIMBLE package library(copula)                  ## call copula package                                  ## Generate bivariate Gamma                                  variates from  Clayton copula theta =3 # the target copula parameter n1=n=3                           ## no. of units used for each                                     treatment C=50 k=6                              ## set up k=6 treatments tU=168∗26                        ## 26 weeks (one week=168                                     hours) for ADT t=seq(0,tU,168∗2)/24             ## check the test each two weeks m=length(t)-1                    ## no. of measurement times dt=t[2:(m+1)]-t[1:m]             ## scaled time increments LL=L=array(dim=c(6,2)) LL[1,]=c(25, 350) LL[2,]=c(45, 650) LL[3,]=c(60, 650) LL[4,]=c(75, 450) LL[5,]=c(75, 550) LL[6,]=c(75, 650) mL1=max(LL[,1]) mL2=max(LL[,2]) L[,1]=(1/(25+273.15)-1/(LL[,1]+273.15))/(1/(25+273.15)-1/ (mL1+273.15)) L[,2]=(log(LL[,2])-log(350))/(log(mL2)-log(350)) r0=-2.902;  r1=0.577; r2=0.533; r3=0.531; be=0.662 v.r=c(r0,r1,r2,r3,be) generatedata=function(v,t,be,n1, theta)  {m=length(t)-1           # no. of measurement times   y=array(dim=c(2,m,n1,k)) for (i in 1:k) {for (j in 1:n1){      cop <- claytonCopula(param = theta, dim = 2)      U <- rCopula(m, cop)      y[1,,j,i]= qgamma(U[,1],shape=v[i]∗dt[i],scale=be)      y[2,,j,i]= qgamma(U[,2],shape=v[i]∗dt[i],scale=be)      } } return(y) # 4 dimensional data } v=rep(0,k)                                                                                                                          # the drift parameter for each treatment v0=exp(r0) for (i in 1:k) {v[i]=v0∗exp(r1∗L[i,1]+r2∗L[i,2]+r3∗L[i,1]                      ∗L[i,2])} #delta yy= generatedata (v,t,be,n1, theta)  # generate the bivariate                                        copula data y1=yy[1,,,]; y2=yy[2,,,] zeros=array(0,dim=c(m,n1,k)) mydata=list(xx=y1, yy=y2,  L=L , r3=0.531) # the data set myinits=function()list(theta=5, r0=-2.8,  r1=0.4, r2=0.4,                         be=0.5) # initial values ClaytonCode=nimbleCode({           # likelihood defined below for (mm in 1:13){ for( nn in 1:n1){   for (ii in 1:6){     delta[mm,nn,ii] <- (exp(r0)∗exp(r1∗L[ii,1]+r2∗L[ii,2]+r3                                     ∗L[ii,1]∗L[ii,2]))∗ 14     dummy[mm,nn,ii] ~ dpois(negLogLike[mm,nn,ii]) # zero trick     u[mm,nn,ii]<- pgamma(xx[mm,nn,ii], shape=delta[mm,nn,ii],                       scale=be)      # marginal gamma CDF     v[mm,nn,ii]<- pgamma(yy[mm,nn,ii], shape=delta[mm,nn,ii],                                        scale=be)     negLogLike[mm,nn,ii]<-  100+(-1)∗logLike[mm,nn,ii]     logLike[mm,nn,ii]<- log(theta+1)+(-theta-1)∗log(u[mm,nn,ii]                                                  ∗v[mm,nn,ii])+           ((-1)∗(2∗theta+1)/theta)∗log(u[mm,nn,ii]ˆ(-theta)+           v[mm,nn,ii]ˆ(-theta)-1)+           log(dgamma(xx[mm,nn,ii], delta[mm,nn,ii],scale=be))+           log(dgamma(yy[mm,nn,ii],delta[mm,nn,ii],scale=be))   ## negLogLike[mm,nn,ii]= (-1)∗log-likelihood, which is      "the negative value of the log-likelihood function"   ## logLike[mm,nn,ii] is the log-likelihood of Eq.(3) }}} theta~ dgamma(.01,.01)    # parameter’s prior distributions r0  ~ dunif(-3.5,-2.5) r1  ~ dgamma(.01,.01) r2  ~ dgamma(.01,.01) be ~  dgamma(.01,.01) }) # end of nimbleCode function                                              # computing the                                              estimation by MCMC                                              method mcmc.out=nimbleMCMC(ClaytonCode,             # call  "Clayton                                              Code" nimbleCode  monitors=c("theta","r0", "r1", "r2","be"),  # interested param-                                                  eters  constants=list(xx=y1, yy=y2, L=L, r3=0.531, n1=n1),  data=list(dummy=zeros),                     # data and constants  inits=myinits, thin=1,                      #  initial values  nchains=2, niter=20000, nburnin=1000,       # 2 chains, 20000                                                MCMC iterations,                                                1000 burn-in periods  summary=TRUE)                               #  summary of the MCMC                                                estimation mcmc.out$summary                             # the MCMC estimates ## The Bayes estimates:  be=0.6242, r0=-2.8598, r1= 0.4528,    r2=0.6806, theta= 2.7733 ## Below are the sample MCMC outputs. #$chain1 #             Mean    Median  St. Dev. 95%CI_low 95%CI_upp #    be      0.6267   0.6231   0.0619    0.5205        0.7647 #    r0     -2.8652  -2.8627   0.1334   -3.1288       -2.6053 #    r1      0.4543   0.4579   0.1364    0.1846        0.7146 #    r2      0.6802   0.6780   0.1255    0.4274        0.9284 #    theta  2.7884   2.7655   0.3917    2.0927        3.6011 #$chain2 #            Mean    Median  St. Dev.  95%CI_low 95%CI_upp #    be      0.6217   0.6183   0.0629    0.5083        0.7606 #    r0     -2.8545  -2.8513   0.1357   -3.1315       -2.6027 #    r1      0.4513   0.4509   0.1303    0.1917        0.7012 #    r2      0.6810   0.6815   0.1218    0.4339        0.9195 #    theta  2.7581   2.7353   0.3877    2.0658        3.5689 #$all.chains #            Mean    Median  St. Dev. 95%CI_low 95%CI_upp #    be      0.6242   0.6205   0.0625    0.5141        0.7623 #    r0     -2.8598  -2.8568   0.1347   -3.1300       -2.6035 #    r1      0.4528   0.4544   0.1334    0.1886        0.7081 #    r2      0.6806   0.6796   0.1237    0.4306        0.9247 #    theta  2.7733   2.7475   0.3900    2.0775        3.5940
Rights and permissions
Copyright information
© 2022 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this chapter
Cite this chapter
Lin, YJ., Tsai, TR., Lio, Y. (2022). Bayesian Estimation for Bivariate Gamma Processes with Copula. In: Lio, Y., Chen, DG., Ng, H.K.T., Tsai, TR. (eds) Bayesian Inference and Computation in Reliability and Survival Analysis. Emerging Topics in Statistics and Biostatistics . Springer, Cham. https://doi.org/10.1007/978-3-030-88658-5_8
Download citation
DOI: https://doi.org/10.1007/978-3-030-88658-5_8
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-88657-8
Online ISBN: 978-3-030-88658-5
eBook Packages: Mathematics and StatisticsMathematics and Statistics (R0)