## Semiparametric Bayesian Analysis Of Survival Data (1996)

Venue: | Journal of the American Statistical Association |

Citations: | 28 - 1 self |

### BibTeX

@ARTICLE{Sinha96semiparametricbayesian,

author = {Debajyoti Sinha and Dipak K. Dey},

title = {Semiparametric Bayesian Analysis Of Survival Data},

journal = {Journal of the American Statistical Association},

year = {1996},

volume = {92},

pages = {1195--1212}

}

### Years of Citing Articles

### OpenURL

### Abstract

this paper are motivated and aimed at analyzing some common types of survival data from different medical studies. We will center our attention to the following topics.

### Citations

9034 | Maximum likelihood from incomplete data via the EM algorithm - Dempster, Laird, et al. - 1977 |

4050 |
Stochastic relaxation, Gibbs distributions, and the Bayesian restoration of images
- Geman, Geman
- 1984
(Show Context)
Citation Context ...bly unknown hyperparameters. Careful applications of some recently popular computational tools, such as posterior likelihood method and the Markov Chain Monte Carlo algorithms such as Gibbs sampling (=-=Geman and Geman, 1984-=-), are used to find posterior estimates of several quantities of interest even when we are dealing with complex models and unusual data structures. The methodologies developed in this paper are motiva... |

2523 |
Equation of State Calculations by Fast Computing Machines, in
- Metropolis, Rosenbluth, et al.
- 1953
(Show Context)
Citation Context ...terior distributions when only conditional distributions are available to sample from. Also this algorithm can be used in conjunction with other Markov chain algorithms, such as Metropolis algorithm (=-=Metropolis et al., 1953-=-), to compute the desired marginal posterior distributions. One can describe the Gibbs sampler algorithm as follows: Given the variables U 1 ; U 2 ; \Delta \Delta \Delta ; U d and an arbitrary set of ... |

2159 |
An Introduction to Probability Theory and its Applications
- Feller
- 1968
(Show Context)
Citation Context ...ed for a stable frailty distribution to model the multivariate survival data. In this model the frailty random variable W follows the positive stable distribution with the Laplace transformation (see =-=Feller, 1971-=-) p(s) = exp(\Gammas ff ) : This is the only frailty model which along with conditional proportional hazards structure of (3.4) gives a proportional hazards structure marginally also. That is, P (T ? ... |

1174 | Bayes factors - KASS, RAFTERY - 1995 |

826 |
Longitudinal data analysis using generalized linear models. Biometrika
- Liang, SL
- 1986
(Show Context)
Citation Context ... Thall also estimates the frailty for each subject instead of a single hyperparameter quantifying the overall variability of the random frailties. Thall and Vail (1990) use the quasi-likelihood idea (=-=Liang and Zeger, 1986-=- and Zeger and Liang,1986) to analyze grouped data, but the method cannot be used to estimate efficiently the parameter of the frailty distribution and the standard error of the treatment effect. One ... |

816 | Inference from iterative simulation using multiple sequences. Statistical Science 7:457–511 - Gelman, Rubin - 1992 |

787 |
A Bayesian Analysis of Some Nonparametric Problems
- FERGUSON
- 1973
(Show Context)
Citation Context ...xample is the h 0 of the Cox's model) of the semiparametric models. For the time being, we are dealing with nonparametric models with no covariates. 5.1 DIRICHLET PROCESS PRIOR The Dirichlet process (=-=Ferguson, 1973-=-) was the first random probability measure used in a nonparametric Bayesian context. It generates a distribution function F (\Delta) on a given sample space, such as the positive real line. It plays t... |

594 | Probabilistic inference using Markov chain Monte Carlo methods - Neal - 1993 |

523 |
Regression Models and Life Tables
- Cox
- 1972
(Show Context)
Citation Context ...tumors for the rats. Markov and semi-Markov models of Lagakos, Sammerand Zelen (1978) often have many parameters and can be hard to interpret. Another class of models - `modulated renewal processes' (=-=Cox, 1972-=- b) - have not been widely used, perhaps because of the difficulty in estimating the asymptotic covariance matrix of the parameter estimates (Oakes, 1981). This difficulty with modulated renewal proce... |

495 |
Analysis of Survival Data
- Cox, Oakes
- 1984
(Show Context)
Citation Context ...(w l t) o exp(fix li ) ; (3:6) where S 0 (\Delta) is the conditional baseline survival function. The random scale model is particularly interesting as it is a conditional accelerated life time model (=-=Cox and Oakes, 1984-=-) given the random effect, where frailty model is a conditional proportional hazards model given the random effect. Unlike in the frailty model the random scale model can introduce negative correlatio... |

419 |
Sampling Based Approaches to Calculating Marginal Densities
- Gelfand, Smith
- 1990
(Show Context)
Citation Context ... draws. It can be shown under rather weak conditions that the Markovian sequence converges to an equilibrium distribution which is the joint distribution of (U 1 ; U 2 ; \Delta \Delta \Delta ; U d ) (=-=Gelfand and Smith, 1990-=-; Tanner 1994). One can use m independent paths of this iterative algorithm to generate m copies of samples from the joint distribution of (U 1 ; U 2 ; \Delta \Delta \Delta ; U d ). After generating m... |

390 | Bayesian computation via the Gibbs sampler and related Markov chain Monte Carlo methods - SMITH, ROBERTS - 1993 |

386 | Marginal Likelihood from the Gibbs Output
- Chib
- 1995
(Show Context)
Citation Context ...erior probabilities have been developed, which include bridge sampling (Meng and Wong, 1993), path sampling (Gelman and Meng, 1994), a data-augmentation-based method to estimate the marginal density (=-=Chib, 1995-=-), ratio importance sampling (Chen and Shao, 1994), and computing Bayes factor with different dimensions of model parameter spaces (Chen and Shao, 1995). A Bayesian survival analyst should adopt the o... |

371 | Variable Selection via Gibbs Sampling - George, McCuloch - 1993 |

216 | Bayesian Model Choice: Asymptotics and Exact Calculations - Gelfand, Dey - 1994 |

210 | Posterior predictive assessment of model fitness via realized discrepancies - Gelman, Meng, et al. - 1996 |

193 | Econometric Analysis.” Second Edition - Greene |

188 | A Monte Carlo implementation of the EM algorithm and the poor man's data augmentation algorithms - WEI, TANNER - 1990 |

182 | Cox’s regression model for counting processes: A large sample - Andersen, Gill - 1982 |

175 | Bayesian Model Choice via Markov Chain Monte Carlo Methods - CARLIN, CHIB - 1995 |

158 | The intrinsic Bayes factor for model selection and prediction - BERGER, PERICCHI - 1996 |

145 | Practical Markov chain Monte Carlo - Geyer - 1992 |

136 | Reversible jump MCMC computation and Bayesian model determination. Biometrika 92:711–732 - Green - 1995 |

124 |
A model for associate in bivariate life tables and its application in epidemiological studies of familial tendency chronic disease incidence
- Clayton
- 1978
(Show Context)
Citation Context ...to be independent and identically distributed for each patient with some parametric distribution with unit mean. The most commonly used frailty distribution is gamma frailty (Oakes, 1982, 1986, 1989; =-=Clayton, 1978-=-, 1991) where the unknown variance of the w l 's (say, ) quantifies the amount of heterogeneity among patients. One important aspect of the model in (3.4) is the distribution of the frailty random var... |

122 |
The impact of heterogeneity in individual frailty on the dynamics of mortality
- Vaupel, Manton, et al.
- 1979
(Show Context)
Citation Context ...ess, the infections within a patient are independent. The given parameters or processes vary over patients according to some distribution. The most widely used conditional model is the frailty model (=-=Vaupel et al., 1979-=-, Aalen, 1988). Let us denote the survival time of the i-th infection to the l-th patient by T li . The l-th patient has been monitored for n l infections. For example, n l = 2 for the kidney infectio... |

115 |
Sampling and Bayes Inference in Scientific Modelling and Robustness
- Box
- 1980
(Show Context)
Citation Context ... and Chib (1995), Green (1994) and Laud and Ibrahim (1994). 8.1 Model Adequacy For model adequacy in a Bayesian framework, much less has been said. The formal Bayesian model adequacy criterion (as in =-=Box, 1980-=-) proposes that the marginal density of the data be evaluated at the observations. Large values support the model, small values do not. In our view, such a checking procedure will be infeasible for th... |

113 | On a class of Bayesian nonparametric estimates: I. Density estimates, The Annals of Statistics 12 - Lo - 1984 |

113 |
Linear Statistical Inference and Its Application. Second Edition
- Rao
- 1973
(Show Context)
Citation Context ...0 (s) (may be unknown) and hyperparameter c as the confidence about the prior mean. The existence of such a process is not at all obvious, since beta variables have cumbersome convolution properties (=-=Rao, 1973-=-). For example, if the hazards in two consequtive intervals h 1 and h 2 are independent beta beta variables, the hazard in the combined interval h = h 1 + (1 \Gamma h 1 )h 2 is not beta. For details, ... |

105 | Model determination using predictive distributions with implementation via sampling-based methods - Gelfand, Dey, et al. - 1992 |

98 | Bayesian “confidence intervals” for the cross-validated smoothing spline - WAHBA - 1983 |

93 | The empirical distribution function with arbitrary grouped, censored and truncated data - Turnbull - 1976 |

91 | Regression analysis of multivariate incomplete failure time data by modeling marginal distributions - WEI, LIN, et al. - 1989 |

90 |
A predictive approach to model selection
- Geisser, Eddy
- 1979
(Show Context)
Citation Context ...ata(i; j); M r ) f(v ij jdata(i; j); M s ) (8:7) or perhaps log d (ij) ; indicates that observation (ij) supports M r over M s . For a scalar summary measure, we turn to the pseudo-Bayes factor PsBF (=-=Geisser and Eddy, 1979-=-) between two models M r and M s , defined as PsBF rs = Q ij d (ij) . Gelfand and Mallick (1995) used the CPO to perform the hazard model selection for right-censored univariate survival data. Ghosh a... |

90 | roughness penalties for probablity densities - Good, &Gaskins - 1971 |

89 |
Nonparametric Bayes estimators based on beta processes in models for life history data
- Hjort
- 1990
(Show Context)
Citation Context ... PRIOR A more general way (other than the way described above) of defining the cumulative hazard is A(t) = Z [0;t] dF (s) 1 \Gamma F (s\Gamma) so that F (t) = 1 \Gamma Y [0;t] f1 \Gamma dA(s)g (5:1) (=-=Hjort, 1990-=-). This coincides with H(t) in the absolutely continuous case and has jumps of size P (T = tjTst) when t is an atom of F . This cumulative hazard rate A(t) can be modeled as a realization of a beta pr... |

85 | Model choice: A minimum posterior predictive loss approach - Gelfand, K - 1998 |

85 |
Bivariate survival models induced by frailties
- Oakes
- 1989
(Show Context)
Citation Context ...pular choice is the gamma distribution, but it is important to note that there are other choices. Another important frailty distribution is the so-called positive stable distribution (Hougaard, 1985; =-=Oakes 1989-=-; Manatunga and Oakes, 1992; Oakes 1992; Oakes 1994). Positive stable distribution has no closed form density function and it is characterized by its Laplace transformation. Stable frailty is the only... |

75 | Predictive Model Selection - Laud, Ibrahim - 1995 |

73 | On use of the EM algorithm for penalized likelihood estimation - Green - 1990 |

63 |
Multivariate generalizations of the proportional hazards model
- Clayton, Cuzick
- 1985
(Show Context)
Citation Context ...he l-th patient has been monitored for n l infections. For example, n l = 2 for the kidney infection data of McGilchrist and Aisbett (1991). The popular Clayton-Cuzick model (Oakes, 1982, 1986, 1989; =-=Clayton and Cuzick, 1985-=- and Clayton, 1991) includes a random effect or frailty (Vaupel et al., 1979) for the group-effect (patient-effect). The conditional hazard function of T li given the unobserved frailty random variabl... |

63 |
Nonparametric Bayesian analysis of survival time data
- Kalbfleisch
- 1978
(Show Context)
Citation Context ...f a prior distribution, generating a distribution F (\Delta) which in turn generates the data. But in survival analysis, modeling of a prior process through a survival function is not very practical (=-=Kalbfleisch, 1978-=-; Burridge, 1981) and in particular the Dirichlet process has some seriously unsettling theoretical consequences (Cornfield and Detre, 1977; Kalbfleisch, 1978). So, in this proposed research full atte... |

61 | Dynamic generalized linear models and Bayesian forecasting (with discussion - West, Harrison, et al. |

48 |
A class of multivariate failure time distributions
- Hougaard
- 1986
(Show Context)
Citation Context ...r which both the conditional hazard and marginal hazard can have the proportional hazards structure of the Cox model. Another characterization of the stable frailty is that it leads to the so-called (=-=Hougaard, 1986-=-) bivariate extreme value distribution for bivariate survival data. More about positive stable frailty will be discussed later. There are some variations of the frailty model of (3.4). Yau and McGilch... |

43 |
A Monte Carlo method for Bayesian inference in frailty models
- Clayton
- 1991
(Show Context)
Citation Context ...tored for n l infections. For example, n l = 2 for the kidney infection data of McGilchrist and Aisbett (1991). The popular Clayton-Cuzick model (Oakes, 1982, 1986, 1989; Clayton and Cuzick, 1985 and =-=Clayton, 1991-=-) includes a random effect or frailty (Vaupel et al., 1979) for the group-effect (patient-effect). The conditional hazard function of T li given the unobserved frailty random variable w l of the l-th ... |

43 | A Bayesian nonparametric approach to reliability - Dykstra, Laud - 1981 |

43 |
A proportional hazards model for interval-censored failure time data
- Finkelstein
- 1986
(Show Context)
Citation Context ...out to compare 48 early breast cancer patients who had been treated with primary radiation therapy and adjuvant chemotherapy to those 46 patients treated with radiotherapy alone (Beadle et al., 1984; =-=Finkelstein, 1986-=-). In this study, a patient who is monitored at some regular intervals (monthly) for cosmetic effect of the therapy may miss some visits and return in a possibly changed state. For example one patient... |

42 | Some covariance models for longitudinal count data with overdispersion - Thall, Vail - 1990 |

41 |
A model for association in bivariate survival data
- Oakes
- 1982
(Show Context)
Citation Context ... l-th patient by T li . The l-th patient has been monitored for n l infections. For example, n l = 2 for the kidney infection data of McGilchrist and Aisbett (1991). The popular Clayton-Cuzick model (=-=Oakes, 1982-=-, 1986, 1989; Clayton and Cuzick, 1985 and Clayton, 1991) includes a random effect or frailty (Vaupel et al., 1979) for the group-effect (patient-effect). The conditional hazard function of T li given... |

39 | Nonparametric Bayesian inference from right censored survival data using the Gibbs sampler - Arjas, Gasbarra - 1994 |