## Sequential Importance Sampling for Nonparametric Bayes Models: The Next Generation (1998)

Venue: | Journal of Statistics |

Citations: | 72 - 5 self |

### BibTeX

@ARTICLE{Maceachern98sequentialimportance,

author = {Steven N. Maceachern and Merlise Clyde and Jun S. Liu},

title = {Sequential Importance Sampling for Nonparametric Bayes Models: The Next Generation},

journal = {Journal of Statistics},

year = {1998},

volume = {27},

pages = {251--267}

}

### Years of Citing Articles

### OpenURL

### Abstract

this paper, we exploit the similarities between the Gibbs sampler and the SIS, bringing over the improvements for Gibbs sampling algorithms to the SIS setting for nonparametric Bayes problems. These improvements result in an improved sampler and help satisfy questions of Diaconis (1995) pertaining to convergence. Such an effort can see wide applications in many other problems related to dynamic systems where the SIS is useful (Berzuini et al. 1996; Liu and Chen 1996). Section 2 describes the specific model that we consider. For illustration we focus discussion on the beta-binomial model, although the methods are applicable to other conjugate families. In Section 3, we describe the first generation of the SIS and Gibbs sampler in this context, and present the necessary conditional distributions upon which the techniques rely. Section 4 describes the alterations that create the second generation techniques, and provides specific algorithms for the model we consider. Section 5 presents a comparison of the techniques on a large set of data. Section 6 provides theory that ensures the proposed methods work and that is generally applicable to many other problems using importance sampling approaches. The final section presents discussion. 2 The Model

### Citations

750 |
A bayesian analysis of some nonparametric problems
- Ferguson
- 1973
(Show Context)
Citation Context ... semi-parametric hierarchical Bayes methods have enjoyed a resurgence of interest with the development of modern Monte Carlo techniques. One popular class of models is based on the Dirichlet process (=-=Ferguson, 1973-=-), and has become known as mixture of Dirichlet process (MDP) models (Antoniak, 1974). The advantage of a MDP model over a standard parametric hierarchical model as in Lindley and Smith (1972) is that... |

488 | Sequential Monte Carlo Methods for Dynamical System
- Liu, Chen
- 1998
(Show Context)
Citation Context ...tisfy questions of Diaconis (1995) pertaining to convergence. Such an effort can see wide applications in many other problems related to dynamic systems where the SIS is useful (Berzuini et al. 1996; =-=Liu and Chen 1996-=-). Section 2 describes the specific model that we consider. For illustration we focus discussion on the beta-binomial model, although the methods are applicable to other conjugate families. In Section... |

434 |
Mixtures of Dirichlet processes with applications to Bayesian nonparametric problems
- Antoniak
- 1974
(Show Context)
Citation Context ...th the development of modern Monte Carlo techniques. One popular class of models is based on the Dirichlet process (Ferguson, 1973), and has become known as mixture of Dirichlet process (MDP) models (=-=Antoniak, 1974-=-). The advantage of a MDP model over a standard parametric hierarchical model as in Lindley and Smith (1972) is that it allows for more flexibility through the incorporation of a nonparametric hierarc... |

424 | Bayesian Density Estimation and Inference Using Mixtures - Escobar, West - 1995 |

283 |
Ferguson distributions via Pólya urn schemes
- Blackwell, MacQueen
- 1973
(Show Context)
Citation Context ... mixing distribution, F jx. We may obtain estimates of all of these quantities on the basis of the simulation techniques that we present. The Polya urn scheme representation of the Dirichlet process (=-=Blackwell and MacQueen, 1973-=-) provides the basis for most modern computational strategies: for exceptions, see Doss (1994), Gelfand and Kuo (1991), and Kuo and Smith (1992). Let the measure ff be written as: ff((\Gamma1; x]) = M... |

198 | Sequential imputation and Bayesian missing data problems - Kong, Liu, et al. - 1994 |

147 | Estimating mixture of dirichlet process models - MacEachern, Mueller - 1998 |

107 |
Estimating Normal Means with a Dirichlet Process
- Escobar
- 1994
(Show Context)
Citation Context ...ities needed to calculate the weights must be evaluated in order to generate values of (s i ; ` i ) in step (A). Finding the weights requires almost no additional computational effort. Gibbs sampler (=-=Escobar, 1994-=-): Each ` i is, in turn, viewed as the last observation (of n) from a Polya urn scheme. The remaining n \Gamma 1 observations form k clusters with locations ` 1 ; : : : ; ` k . A cluster membership an... |

106 | Estimating normal means with a conjugate style dirichlet process prior - MacEachern - 1994 |

101 | A semiparametric Bayesian model for randomised block design - Bush, MacEachern - 1996 |

82 | The collapsed Gibbs sampler in Bayesian computations with applications to a gene regulation problem - Liu - 1994 |

73 | Bayes Estimates for Linear Model (with discussion - Lindley, Smith - 1972 |

52 |
Nonparametric hierarchical Bayes via sequential imputations
- Liu
- 1996
(Show Context)
Citation Context ...tly the effective sample size ESS = R=(1 + var(w)). In the sequel, we estimate ESS by replacing var(w) withsvar(w ). The following scheme was proposed in Liu (1996): Sequential Importance Sampler S1 (=-=Liu, 1996-=-): Repeat steps (A) and (B) for i = 1; : : : ; n. A Generate (s i ; ` i )j(s !i ; ` !i ; x !i ; x i ) by first generating s i from the distribution in (5). If s isk, then set ` i = ` s i . If s i = k+... |

36 | Bayesian nonparametric estimation for incomplete data via successive substitution sampling - DOSS - 1994 |

27 | Nonparametric Bayesian bioassay including ordered polytomous response - Gelfand, Kuo - 1991 |

19 |
Empirical Bayes estimation of a binomial parameter via mixtures of Dirichlet processes
- Berry, Christensen
- 1979
(Show Context)
Citation Context ...on of ` into clusters, and the number of components in the mixture grows exponentially. Because of the analytic intractability of this type of model, early investigators developed approximations (see =-=Berry and Christensen, 1979-=-) or simulation methods (see Kuo, 1986). For both the SIS and the Gibbs sampler, the key calculations stem from the Polya urn representation of the Dirichlet process: Pr(s i = j j s !i ; ` !i ; x !i ;... |

15 | Bayesian computations in survival models via the Gibbs sampler - Kuo, Smith - 1992 |

9 | Spectral analysis for discrete longitudinal data - Beckett, Diaconis - 1994 |

5 |
Dynamic graphical models and Markov chain Monte Carlo methods
- Berzuini, Best, et al.
- 1994
(Show Context)
Citation Context ...ed sampler and help satisfy questions of Diaconis (1995) pertaining to convergence. Such an effort can see wide applications in many other problems related to dynamic systems where the SIS is useful (=-=Berzuini et al. 1996-=-; Liu and Chen 1996). Section 2 describes the specific model that we consider. For illustration we focus discussion on the beta-binomial model, although the methods are applicable to other conjugate f... |

1 | Unpublished - Gopalan - 1994 |

1 | Discussion of "Bayesian computations in survival models via the Gibbs sampler" by Kuo and Smith - MacEachern - 1992 |