## Using simulation methods for Bayesian econometric models: Inference, development and communication (1999)

### Cached

### Download Links

Venue: | Econometric Review |

Citations: | 270 - 15 self |

### BibTeX

@ARTICLE{Geweke99usingsimulation,

author = {John Geweke},

title = {Using simulation methods for Bayesian econometric models: Inference, development and communication},

journal = {Econometric Review},

year = {1999},

pages = {1--126}

}

### Years of Citing Articles

### OpenURL

### Abstract

This paper surveys the fundamental principles of subjective Bayesian inference in econometrics and the implementation of those principles using posterior simulation methods. The emphasis is on the combination of models and the development of predictive distributions. Moving beyond conditioning on a fixed number of completely specified models, the paper introduces subjective Bayesian tools for formal comparison of these models with as yet incompletely specified models. The paper then shows how posterior simulators can facilitate communication between investigators (for example, econometricians) on the one hand and remote clients (for example, decision makers) on the other, enabling clients to vary the prior distributions and functions of interest employed by investigators. A theme of the paper is the practicality of subjective Bayesian methods. To this end, the paper describes publicly available software for Bayesian inference, model development, and communication and provides illustrations using two simple econometric models. *This paper was originally prepared for the Australasian meetings of the Econometric Society in Melbourne, Australia,

### Citations

4100 | Stochastic relaxation, Gibbs distributions, and the Bayesian restoration of images - Geman, Geman - 1984 |

2564 | Equation of state calculations by fast computing machines - Metropolis, Rosenbluth, et al. - 1953 |

1546 | Bayesian Data Analysis - Gelman, Carlin, et al. - 1997 |

1429 | Statistical Decision Theory and Bayesian Analysis - Berger - 1985 |

1385 | Monte Carlo Sampling Methods Using Markov Chains and Their - Hastings - 1970 |

1205 | Bayes factors - Kass, Raftery - 1995 |

1194 |
Bayesian Theory
- Bernardo, Smith
- 2000
(Show Context)
Citation Context ... ~N β, j j , , H− 1 12K conditional on a known value of the j ( ) = ( ) disturbance precision h. A corresponding sequence of kernels is ∫ Θ Θ . 9 This follows from the Neyman factorization criterion (=-=Bernardo and Smith, 1994-=-, Section 4.5.2). Less formally, from (2.2.1), it is clear that one only needs to know s T to write the likelihood function for β and h. 10* p ( β Aj)= exp −( )( β −β) j β β ′ ⎡ −1 ⎤ * * ⎢ 12 H ( − )... |

860 | Markov chains for exploring posterior distributions
- Tierney
- 1994
(Show Context)
Citation Context ...m of the algorithm is due to Hastings (1970). The Metropolis et al. (1953) form takes q θ m ( ) ,θ * ( )= q θ * ,θ m ( ) ( ). A simple variant that is often useful is the ( m) * * independence chain (=-=Tierney, 1994-=-), whereby q ( θ , θ )= k( θ ). Then ( ) ( ) m ⎧ ( ) θ θ θ m T A ( ) ⎪ ⎫⎪ αθ ( θ )= ⎨ ( m) ⎬ ( m) ⎩⎪ ( θ T A) ( θ ) ⎭⎪ θ = p , k ⎧⎪ ⎫⎪ , min , min⎨ ⎬ p , k ⎩⎪ ( ) ⎭⎪ w * * Y * 1 , 1 , * Y w where wθ p... |

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

706 |
Essays in Positive Economics
- Friedman
- 1953
(Show Context)
Citation Context ...d in the subsequent literature. The decomposition is the quantitative expression of the fact that predictive power is the scientifically relevant test of the validity of a hypothesis (Jeffreys, 1939; =-=Friedman, 1953-=-). This review concentrates entirely on exact, finite sample methods. As is the case in non-Bayesian statistics, given suitable regularity conditions, there exist useful asymptotic approximations to t... |

659 |
The Calculation of Posterior Distributions by Data Augmentation. With Discussion and With a Reply by the Authors
- Tanner, Wong
- 1987
(Show Context)
Citation Context ...duced by Geman and Geman (1984). This version was subsequently shown to have great potential for Bayesian computation by Gelfand and Smith (1990). Their work, combined with data augmentation methods (=-=Tanner and Wong, 1987-=-), has proven very successful in the treatment of latent variables in econometrics. Since 1990, application of MCMC methods has grown rapidly (Chib and Greenberg, 1996). This section and the next conc... |

525 | Bayesian analysis of binary polychotomous response data - Albert, Chib - 1993 |

502 | Understanding the Metropolis-Hastings algorithm
- Chib, Greenberg
- 1995
(Show Context)
Citation Context ...s lower (one). There is a simple two-step argument that motivates the convergence of the sequence θ m ( ) { } generated by the Hastings-Metropolis algorithm to the posterior. (This approach is due to =-=Chib and Greenberg, 1995-=-.) First, observe that if the transition probability function ( ) satisfies the reversibility condition p θ m ( ) ,θ m+1 ( ) ( ) ( )p θ m ( ) ,θ m+1 ( ) ( )= p θ m+1 ( ) ( )p θ m+1 ( ) ,θ m ( ) (3.4.2... |

478 | Monte Carlo Methods - Hammersley, Handscomb - 1964 |

471 | Stochastic Simulation - Ripley - 1987 |

399 | Marginal likelihood from the Gibbs output - Chib - 1995 |

370 |
Bayesian Inference in Econometric Models Using Monte Carlo Integration.Econometrica
- Geweke
- 1989
(Show Context)
Citation Context ...equacy of j( θ) in all other cases. The ratio var g θ, YT YT, σ A 2 [ ( ) ] has been termed the relative numerical efficiency (RNE) of the importance sampling approximation to Eg YT, YT, A θ [ ( ) ] (=-=Geweke, 1989-=-b): it indicates the ratio of iterations using p ( θYT, A) itself as the importance sampling density, to the number using j( θ), required to achieve the same accuracy of approximation of g . Since bot... |

355 | Evaluating the accuracy of sampling-based approaches to calculating posterior moments - Geweke - 1992 |

326 | Applied statistical decision theory - Raiffa, Schlaifer - 1961 |

299 | Numerical recipes in FORTRAN: the art of scientific computing - Press, Teukolsky, et al. - 1992 |

269 |
Multiple Time Series
- Hannan
- 1970
(Show Context)
Citation Context ...1992) and Geyer (1992) and their discussants. Geweke (1992) developed a consistent estimator of σ 2 in (3.7.1) under the strong condition that conventional time series mixing conditions (for example, =-=Hannan, 1970-=-, 207–210) apply to θ m ( ) { }. There is no analytical foundation for this assumption, but these methods are now widely used and have proven reliable in the sense that they predict well the behavior ... |

268 | An Introduction to Bayesian Inference in Econometrics - Zellner - 1971 |

223 | Bayesian Model Choice: Asymptotics and Exact Calculations - GELFAND, DEY - 1994 |

157 | Efficient Simulation from the Multivariate Normal Distribution Subject to Linear Inequality Constraints and the Evaluation of Constraint Probabilities - Geweke - 1989 |

120 | Sampling and Bayes’ inference in scientific modelling and robustness with discussion - Box |

119 | Markov Chain Monte Carlo Simulation Methods in Econometrics,” Econometric Theory
- Chib, Greenberg
- 1996
(Show Context)
Citation Context ...ned with data augmentation methods (Tanner and Wong, 1987), has proven very successful in the treatment of latent variables in econometrics. Since 1990, application of MCMC methods has grown rapidly (=-=Chib and Greenberg, 1996-=-). This section and the next concentrate on a heuristic development of two widely used variants of these methods, the Gibbs sampler and the Hastings-Metropolis algorithm. The general theory of converg... |

96 | The Structure of Scientific Revolutions. Second edition - Kuhn - 1962 |

93 |
Generalized linear models with random effects: A Gibbs sampling approach
- Zeger, Karim
- 1991
(Show Context)
Citation Context ...ampling, although this process is transparent in most software (Geweke, 1996, Section 2). Another such strategy that is quite useful in Bayesian econometrics is the Metropolis within Gibbs algorithm (=-=Zeger and Karim, 1991-=-; Chib and Greenberg, 1996). In a two-block Gibbs sampler, suppose that it is straightforward to sample from p θ YT, θ , A , but the distribution corresponding to ( ( 2) ( 1) ) ( () 1 ( 2) ) p θ YT, θ... |

81 | Theory of Probability, Third Edition - Jeffreys - 1961 |

65 | Monte Carlo Simulation and Numerical Integration
- Geweke
- 1996
(Show Context)
Citation Context ...egy to the conditional distribution. For example, draws from a multivariate transition density entail both conditioning and acceptance sampling, although this process is transparent in most software (=-=Geweke, 1996-=-, Section 2). Another such strategy that is quite useful in Bayesian econometrics is the Metropolis within Gibbs algorithm (Zeger and Karim, 1991; Chib and Greenberg, 1996). In a two-block Gibbs sampl... |

64 | Bayesian estimates of equation system parameters: An application of integration by Monte Carlo - Kloek, Dijk - 1978 |

52 |
Antithetic Acceleration of Monte Carlo Integration in Bayesian Inference
- Geweke
- 1988
(Show Context)
Citation Context ...on requires the solution of a system of nonlinear differential equations. 15In some cases, the left side of (3.0.2) can be made quite small indeed, and asymptotically it may be made to approach zero (=-=Geweke, 1988-=-). 18∑ M ( ) ( ) M ( ) ( ) ( ) ( )⎯ → m= 1 ∑m= 1 [ T ] m m m as .. wθ h ω wθ E h ω Y , A. ⎯ ( ) The ability to generate such sequences has improved greatly in the past ten years, due in large part to... |

52 | Optimum Monte-Carlo sampling using Markov chains - Peskun - 1973 |

38 |
Exact predictive densities for linear models with ARCH disturbances
- Geweke
- 1989
(Show Context)
Citation Context ...equacy of j( θ) in all other cases. The ratio var g θ, YT YT, σ A 2 [ ( ) ] has been termed the relative numerical efficiency (RNE) of the importance sampling approximation to Eg YT, YT, A θ [ ( ) ] (=-=Geweke, 1989-=-b): it indicates the ratio of iterations using p ( θYT, A) itself as the importance sampling density, to the number using j( θ), required to achieve the same accuracy of approximation of g . Since bot... |

35 | Semiparametric Estimation of a Hedonic Price Function - Anglin, Gencay - 1996 |

24 | Hypothesis testing and model selection via posterior simulation - Raftery - 1996 |

17 | Simple Conditions for the Convergence - Roberts, Smith - 1994 |

16 | The Central Limit Theorem for Markov Processes - Cogburn - 1972 |

16 | Bayesian statisticians and remote clients - Hildreth - 1963 |

16 | Testing normality in econometric models - Kiefer, Salmon - 1983 |

15 | Practical Markov Chain Monte Carlo. Statistical Science. 7:473-511 Maataloustieteen Päivät 2006. www.smts.fi 7 - Geyer - 1992 |

12 | Frequentists and Subjectivist Perspectives on the Problems of Model Building in Economics - Poirier |

9 | Mixture of normals probit models - Geweke, Keane - 1999 |

8 | The price sealed bid auction with asymmetric bidders: theory and applications - Bajari - 1997 |

5 |
An Empirical Analysis of Income Dynamics Among
- Geweke, Keane
- 1997
(Show Context)
Citation Context ...ge and education, it is natural to consider reasonable ranges for earnings ratios at different age and education levels and nearly impossible to think about individual coefficients of the polynomial (=-=Geweke and Keane, 1997-=-). Moreover, formulation of prior distributions over the substantive, model invariant elements of ω provides considerable discipline in developing prior distributions ( Aj) that are at least reasonabl... |

4 | Comparing and Choosing Between Two Models with a Third Model in the Background - Poirier - 1997 |

3 | Bayesian Comparisons of Simple Macroeconomic Models - Geisel - 1973 |

3 |
Bayesian multivariate exogeneity analysis: An application to a UK money demand equation
- Steel, Richard
- 1991
(Show Context)
Citation Context ...still pertains. For further discussion of ancillarity, see Bernardo and Smith (1994, Section 5.1.4). The condition of weak exogeneity in the econometrics literature (Engle, Hendry, and Richard, 1983; =-=Steel and Richard, 1991-=-) is closely related. 6 More generally, the precision of any random variable is the inverse of its variance. 6and (2.1.9) s 2 h ~ χ 2( ν), where H is a fixed precision matrix, β is a fixed mean vecto... |

2 | Theory of probability, first edition - Jeffreys - 1939 |

2 | Bayesian Diagnostic Testing in the General Linear Normal Regression Model. Bayesian Statistics 3 - Poirier - 1988 |