ESTIl\1ATING BAYES FACTORS VIA POSTERIOR SIMULATION \VITH THE LAPLACE-lVIETROPOLIS ESTIlVIATOR
|
|
- Lionel Washington
- 5 years ago
- Views:
Transcription
1 ESTIl\1ATING BAYES FACTORS VIA POSTERIOR SIMULATION \VITH THE LAPLACE-lVIETROPOLIS ESTIlVIATOR by Steven M. Lewis Adrian E. Raftery TECHNICAL REPORT No. 279 November 1994 Department of Statistics, GN-22 University of Washington Seattle, Washington USA
2 Estin1ating Bayes Factors via Posterior Simulation with the Laplace-Metropolis Estin1ator Steven 1\1. Lewis and Adrian E. Raftery University of vvashington November 30, 1994 Abstract The key quantity needed for Bayesian hypothesis testing and model selection is the marginal likelihood for a model, also known as the integrated likelihood, or the marginal probability of the data. In this paper we describe a way to use posterior simulation output to estimate marginal likelihoods. vve describe the basic Laplace Metropolis estimator for models without random effects. For models with random effects the compound Laplace-Metropolis estimator is introduced. This estimator is applied to data from the World Fertility Survey and shown to give accurate results. Batching of simulation output is used to assess the uncertainty involved in using the compound Laplace-Metropolis estimator. The method allows us to test for the effects of independent variables in a random effects model, and also to test for the presence of random effects. \VORDS: Laplace-Metropolis estimator; Random effects models; Marginal likelihoods; l-'"cte.y"u,, Cll111ll1(],'ulVJeJ, World Survey. 1 Introduction Bayesian solution to moael sell,;ction pf()bleii1ts IS
3 rallo B Ol, for comparing model agiun:st model 1\J 1 ivia agamsl to pnor nh"p, vp d data, Y, In IS ratio integrated, likelihoods two models being compared. Hence calculation of Bayes factors boils down to marginal likelihoods, where () m is the vector of parameters in model AIm and! (() m I j\jm) is its prior density. Dropping the notational dependence on the model, this can be rewritten as!(y) = j!(y I ())!(()) d(). (1) Historically the integration required for calculating marginal likelihoods has been done taking advantage of conjugacy or by assuming approximate posterior normality. In other cases requisite integrals have been approximated using such methods as Gaussian quadrature aylor and Smith 1982), the Laplace approximation (de Bruijn 1970; Tierney and Kadane 1986) or Monte Carlo methods. With the availability of increasing computer power, Markov chain Monte Carlo (MCMC) has become a reasonable alternative. have been a number of alternative ways suggested to use posterior simulation output, as that produced by MCMC, to estimate marginal likelihoods. These are Raftery (1995) and include importance sampling methods as calculating TTnnn,r mean of the output likelihoods, vanous ways of combining prior and such as paper we (le:scflbe CH1'vullvl way to use post(~n()r 0J,1l1l,llCLCIVll n"t"l1t
4 for hi H'-,rrlc,ir-, I IJ<l.Pl<l.LC method at two ditter ent D:loljels. from apply LJGlVl'CLLC In(~tJJlod is applied at a second to integrate out each of the random effect parameters. [n Section 6 the compound Laplace-Metropolis estimator is used to calculate log marginal likelihoods for a number of different models fit to data collected in Iran as part of the World Survey. By taking the difference between log marginal likelihoods for two competing models we are able to estimate the Bayes factor for comparison of the two models. How are log marginal likelihoods estimates produced the compound In Section 7 we demonstrate a method to adapt one way assessmg variability of Monte Carlo estimators, namely the batching of simulation output, for the purpose of assessing the variability of the compound Laplace-Metropolis estimator. We close by comparing the estimates of the log marginal likelihoods provided by the compound Laplace-Metropolis estimator against a "gold standard" determined on the basis Dvlre.yy,.clu large samples from the joint prior distribution of all parameters. The compound Laplace-Metropolis estimates are shown to be remarkably accurate, particularly in light of substantially reduced computing time required. 2 Calculating the Marginal Likelihood Using the Prior Distribution Before describing the Laplace-Metropolis estimator, we first describe a method for obtaining for compound marginal likelihoods we can compare the ap1prc)xlitlate ll not lt1 can paper..gel]eral it is
5 one be<:on1es a good vjvhhlcv'jl to 1'1'>,'11,',0. mt(:;gr,'tl m eqllatlon (1). In error to an it was necessary to use a sample roughly.50 million from the prior The computing effort needed to obtain such a sample can be enormous, days of computer time on most workstations currently available. In this paper will describe an estimator which may be used to estimate the logarithm of the integral in equation (1) in substantially less computer time. However, as a way to check accuracy of the estimator which we describe, we have also calculated Monte Carlo estimate using equation (2) for some of the models demonstrated in this paper; we will refer to this Monte Carlo estimate as the "actual" log marginal likelihood. 3 The Laplace-Metropolis Estimator of the Marginal Likelihood Raftery (1995) has recently proposed a method for estimating the marginal likelihood by combining the Laplace approximation with MCMC. He suggests using the Metropolis al- (Metropolis et as a means for estimating quantities for La])l;:Lce approximation. By applying the Laplace method we can derive the following approximation for the marginal likelihood: P 1 j. r.y) :::::; (2JrI2"IH*I'i j'(8*') f(y 18*). \ J \ J II. J I I (3) 8* is value of 8 which h (8) == log {j (8) j (Y I 8)} attains its maximum, H* 1Tlv',o.r<,,o. Hessian h evaluated at 8* P is dimension of the parameter nurfi(~n;::al reasons IS to log likelihoods, it as (Y P 2 {2Jr {IH* {f Y 8*)}.
6 estimate 0* be to compute h (0) h (0) is large:;t we consider of 0 which minimizes from 0*. can a of computing effort. multivariate median, or L 1 center, which is defined as J l: 10 ( ) - OU) I, =1 denotes L 1 distance. \Ve use this as an estimator of the posterior mode. 'I'he other quantity needed for the Laplace-JVletropolis estimator is H*. This is asympequal to the posterior variance matrix. \Ve could use the sample covariance matrix of the simulation output for H*. However, since MCMC trajectories are known to take occasional distant excursions, it would probably be a good idea to use a robust estimator for the posterior variance matrix. One such estimator is the weighted variance matrix estimate with weights based on the minimum volume ellipsoid estimate of Rousseeuw and van Zomeren (1990). 4 The Compound Laplace-Metropolis Estimator for Hierarchical Models recent work, to be described shortly, we have found need of marginal likelihoods for a UUHf',>, of alternative hierarchical, i.e. random effects, models. Here we use the term hierarmodel in the sense used by Raudenbush (1988), observations are made on subjects which themselves should for a model in which several grouped into higher level enorder to explicitly take into account aggregation m sampling design. It ]S the case that hierarchical involve a very large number nuisance paramrandom effects, as a result calculating marginal llk,ell1]o()(1s becojme~s 00.-_.. ",; assllmptlon IS quiestloilat)le.
7 lul'hhl\.a> pal['arnel;ers, oc, rest of 8 (fj, term in the Laplace-Metropolis estimator, equation (,requiring additional attention a random effects model is the log-likelihood, log {f (Y! 8*)}. Calculation of the terms in equation can be accomplished precisely as already described in Section 3. We still need to find the posterior mode of 8. To do so we locate the L 1 center of (fj, E. The posterior variance of the fixed effects, fj, can still be used for H*. The term IS logarithm joint prior density of fixed parameters and variance hyperparameter. many random effects models the random effects are assumed to conditionally independent given the other parameters in the model, such as the fixed effect parameters and the hyperparameters. vve can take advantage of this assumption to calculate the loglikelihood as simply the sum of the log-likelihoods for each of the random effects. In other words, log {f (Y I fj, E)} I L log {f (Yi I fj, E)}, i=l (5) where log {f (Y i I fj, En (6) CYi is the random effect parameter for the i th context. Equation holds for any (fj, E). In particular, it holds for (0*), the joint mode of IbtltH)n. f (Y i I ai, fj, a result we can each of log- equation conditioning on (0*). integrals on For
8 to as a compound Lapliicc:-lVletrolDoJllS C;",lJllCLvUl IS ;\.1 P 2 1 {2Ti1+ J 2 IH*I} + log (fj*} I Lii i=1 5 The Laplace-Metropolis Estimator for the Logistic Hierarchical Model In this section we derive the Laplace-Metropolis estimator for one of the most frequently of hierarchical model, namely logistic hierarchical model. Here we assume the data are produced by a mixed logistic model. That is, we assume logit(1r) = X'T7 + a, (8) where the data may take on only values 0 or 1, 1r = {Tiit}, Tiit is the probability that the tth observation within the i th random effect is a 1, and X is a matrix of covariate information. The likelihood of the vector of observations for the i th random effect, Y i, is j 'ry! \ \ i 1'T7,0'i) = If we also assume the prior distribution of each of the random effect parameters IS Gaussian with mean 0 and variance L; and independent of the fixed effect parameters the other random effect parameters, it is straightforward to calculate the conditional likelihood of the random effect's vector of observations as 1 \ ~ (2Tit) exp {hz )} + 'X 1 itt! } }
9 can locate the mode we UCllUI,C a iterations the second we square root of C1elLerrmrla;llt of minus the inverse of the Hessian conditional on (0*) evaluated at, I IS vve now have all the quantities needed to use the Laplace method to approximate the integrals on the right-hand side of equation (6). Doing so, we find that a Laplace estimate for random effect log conditional likelihood is L; = These Laplace estimates of the log conditional likelihoods may then be used in equation (7) to calculate a compound Laplace-Metropolis estimate for hierarchical models. In the next section we will show how this works for an example using data collected in Iran as part of the 'World Fertility Survey. 6 Example Using Data from the World Fertility Survey The Iran Fertility Survey (IFS) was a part of the 'World Fertility Survey CWFS). There is m literature describing the details WFS. volume edited serves as a ralldc)ml primary summary publication on the \:VFS. The covariate m10rjmajlcm as
10 m was wuelller or not a woman expene:ncea a m a we to as exposure-years. model consisted of 4 fixed effect parameters in addition to the intercept. The 4 fixed were the age of the woman during each exposure-year (centered at the average li1 IFS data, an indicator variable which is 1 for the first exposure-year of the interval and 0 otherwise, the woman's parity (number of previous children born) during each exposure-year the woman's completed education level (a 6 level categorical variable). exanlpje we also included a random parameter for each woman in the sample. llnple~mient;eda Metropolis algorithm for estimating parameters of a mixed logistic, in a Fortran program written specifically to handle event history data sets (Lewis 1993, 1994; Raftery, Lewis and Aghajanian 1994). \lve obtained the results shown in Table 1. Metropolis was run for a total of 5,500 iterations of which the first 500 were discarded for "burn-in". Table 1: RegT'ession pammeter' estimates fot' a sample ft'om the IFS event histot'y data set llsing the 4 fixed effeet model. Intercept Centered age First of interval nr,cn"",,,,,,, section
11 next sectlo,n we one way doing this. 7 Assessing variance of the estimator using hatching assess the the Laplace-Ivletropolis estimator we adapted one of the most used methods for assessing the uncertainty involved in Monte Carlo estimation, the method of batch means, to the situation at hand. The use of this method for ca.lculating the Monte Carlo variance for the mean of a functional of posterior simulation output has been discussed numerous authors, including Hastings (1970), Schmeiser (1982) and (1992). The underlying idea is to divide an entire MCMC sample into a fairly small number of batches, say B, of equal size. The mean of the simulations within each batch is found, producing B estimates of the mean. Under relatively mild conditions these B estimates will be essentially independent. The sample of the B estimates can then be used to provide a compound estimate of the mean along with an estimate for its variance. To apply this idea to assessing the uncertainty of the Laplace-Metropolis estimator we can ca.lculate separate Laplace-Metropolis estimates within each of B batches, take the mean of B estimates as a compound Laplace-Metropolis estimate and take the variance the mean as an estimate of the variance of the compound Laplace-Metropolis estimate. In other batch we can find Laplace-Metropolis estimate for the log marginal likelihood use Mb = ~ log {2r.} + ~ log {IH*I} + log {f (O*)} +t {i "=1 B estimates to calculate an overall Laplace-Metropolis estimate, (9) A1 Hlt~LHOU to
12 at some if is not or is not sampling from stationary distribution, this is usually apparent sequential plots of one or more fixed effect realizations. The sequential plot of the intercept realizations is the plot which most often exhibits difficulties in the Markov chain. Figure 1 shows the sequential realizations of intercept parameter for the 4 fixed effects model The sequential plots of fixed effects were similar to the intercept plot. In this case the Markov chain seems c;ilvul',h and is likely to be sampling from stationary distribution. N o N I L... j '-,J r , I o : realizatzons of 'UlJ' P'l"J'/-"i'jf pararneter. It is interesting to look at eslt!n1a1;ed marginal denslth~s for some random parameters.
13 erroneous a such as ls pra,ctlce, is quite likely to lead to l!) 0 -.:::t 0 Ct') 0 C\I 0 1,... I Figure 2: Estimated marginal distribution of the intercept parameter. Table 2 shows the estimated log marginal likelihood within each batch. The first batch consisted of the 501 st through 5, sooth iterations. The second batch was made up of it 501 through 10,500, and so forth. Shown are the contribution of the within batch maximized log-likelihood, the contribution of the other three terms of equation to each within batch log marginal likelihood and in the rightmost column the within batch IIli:lTf'"IIli'H likelihood; predominant contribution of the maximized log-likelihood is
14 /Yxample of calculating Laplace-1Vletropolis estimates for 8PIJrtrrtU batches using the SZTliWluum output for the /t effects model., Batch \Vithin Batch ~Within Batch I \Vithin Batch Loglikelihood Other I Logp~terior 'Lf=l i Terms ;\/h I I I H Discussion Raftery (1995) originally proposed Laplace-Metropolis estimator to get around limitations encountered when trying to use the Laplace method. He also proposed use of the L 1 center as an approximation for the posterior mode in the situation where it is impractical,olllluiclv'-u parameter vector. A 1'0,.1101'
15 we COJ]lD,OUnd Lapi,1,CEHVlet,ropo!ls e:3tllnator are reiilajl"ka.ui) approximations obtained accurate. standard solution for comparing two models is to compute Bayes factor. The Bayes factor is ratio of two marginal likelihoods. In this paper we demonstrated how we were able to obtain good approximations for marginal likelihoods in hierarchical models. \Ve found a point estimate of for the log marginal likelihood an example model with 4 fixed effects. \Ve have also fit a number of other mcloels to IFS event history set. example, when we add husband's level completed education to the model, we get an approximate log marginal likelihood of Hence, Bayes factor found for comparing the model without husband's education against the model with husband's education is approximately e 3. 0 ~ 20, providing evidence for the smaller model. Models containing other potential covariates may be similarly compared. Bayes factors can be used not only to compare models containing different fixed effect parameters but also to assess whether the data provide evidence for or against the presence of random effects. In Section 7 we found that a compound Laplace-Metropolis estimate of the log marginal likelihood for the 4 fixed effects model was This model included a woman in the sample. Should we have included rallg<)m 'AL"A.\.,,~ in the model? If we can calculate the log marginal likelihood of a model without random effects, we will be able to determine a Bayes factor for comparing the model with random effects to the model without random effects. The marginal likelihood for the model without random effects can be approximated using Laplace method (Raftery 1993). For the 4 fixed effects model this was Hence the Bayes factor for comparing the model with random effects against HIe model without GH1'UVJ.ll ettelcts is exp { } ~ There is some weak the model incorporating random effects. References
16 C.,1. -Tra<:tH:;al Markov " Statistical,.",JCU:!ICe I, 1- (1970). "Monte sampling methods using Markov chains and their applications. Biometrika 57, 97~109. Kass, R. and Raftery, A.E. (1995). "Bayes factors. Jour'nal of the American Statistical Association, To appear. Lewis, S.M. (1993), "Contribution to the discussion of three papers on Gibbs sampling and related Markov chain Monte Carlo methods." ser'ies B, 55, 79~81. S.M. (1 lvlonte Carlo 111ethods. ivlodeling of Discrete Unpublished doctoral University of Washington, Seattle, Wa. Meng, X.I... and 'Wong, vv.h. (1993). Journal of the Royal Statistical Society! History Data Using Markov dissertation, Department of Statistics, "Simulating ratios of normalizing constants VIa a simple identity." Technical Report no. 365, Department of Statistics, University of Chicago. Metropolis, N., Rosenbluth, A.vV., Rosenbluth, M.N., Teller, A.H. and Teller, E. (1953). "Equations of state calculations by fast computing machines." Physics 21, l087~1092. Journal of Chemical Naylor, J.C. and Smith, A.F.M. (1982). "Applications of a method for the efficient computation of posterior distributions." Applied Statistics 31, 214~225. N M.A. and Raftery, A.E. (1994). "Approximate Bayesian inference by the weighted likelihood bootstrap (with discussion)." Journal of the Royal Statistical Society! Series B 56, 3~48. A.E. (1993). "Approximate Bayes factors and accounting for model uncertainty models." no. 255, of :::itcltistics. D.J. :::ipleg;elhalter
17 A. S.M., A. Kahn, M.J. (,,,roup'u data." \Vorking Paper No. In UE~rrllog]['aT)ny and Ecology, of <Washington, JC'tbLLIC, S.VV. (1988). "Educational applications of hierarchical linear models: A re Journal of Educational Statistics 13, 85~116. Ilousseeuw, P.J. and van Zomeren, B.C. (1990). "Unmasking multivariate outliers and discussion). Journal the American Statistical Association 85, 6:3: Schmeiser, B. (1982). "Batch size effects in the analysis of simulation output." Operations Research 30, 556~568. Terrell, C.R. (1990). "The maximal smoothing principle in density estimation." Journal of the American Statistical Association 85, 470~477. Tierney, L. and Kadane, J.B. (1986). "Accurate approximations for posterior moments and marginal densities." Journal of the American Statistical Association 81,
A note on Reversible Jump Markov Chain Monte Carlo
A note on Reversible Jump Markov Chain Monte Carlo Hedibert Freitas Lopes Graduate School of Business The University of Chicago 5807 South Woodlawn Avenue Chicago, Illinois 60637 February, 1st 2006 1 Introduction
More informationMultilevel Statistical Models: 3 rd edition, 2003 Contents
Multilevel Statistical Models: 3 rd edition, 2003 Contents Preface Acknowledgements Notation Two and three level models. A general classification notation and diagram Glossary Chapter 1 An introduction
More informationMarkov Chain Monte Carlo methods
Markov Chain Monte Carlo methods Tomas McKelvey and Lennart Svensson Signal Processing Group Department of Signals and Systems Chalmers University of Technology, Sweden November 26, 2012 Today s learning
More informationBagging During Markov Chain Monte Carlo for Smoother Predictions
Bagging During Markov Chain Monte Carlo for Smoother Predictions Herbert K. H. Lee University of California, Santa Cruz Abstract: Making good predictions from noisy data is a challenging problem. Methods
More informationSession 3A: Markov chain Monte Carlo (MCMC)
Session 3A: Markov chain Monte Carlo (MCMC) John Geweke Bayesian Econometrics and its Applications August 15, 2012 ohn Geweke Bayesian Econometrics and its Session Applications 3A: Markov () chain Monte
More informationPrinciples of Bayesian Inference
Principles of Bayesian Inference Sudipto Banerjee University of Minnesota July 20th, 2008 1 Bayesian Principles Classical statistics: model parameters are fixed and unknown. A Bayesian thinks of parameters
More informationST 740: Markov Chain Monte Carlo
ST 740: Markov Chain Monte Carlo Alyson Wilson Department of Statistics North Carolina State University October 14, 2012 A. Wilson (NCSU Stsatistics) MCMC October 14, 2012 1 / 20 Convergence Diagnostics:
More informationMonte Carlo in Bayesian Statistics
Monte Carlo in Bayesian Statistics Matthew Thomas SAMBa - University of Bath m.l.thomas@bath.ac.uk December 4, 2014 Matthew Thomas (SAMBa) Monte Carlo in Bayesian Statistics December 4, 2014 1 / 16 Overview
More informationDoing Bayesian Integrals
ASTR509-13 Doing Bayesian Integrals The Reverend Thomas Bayes (c.1702 1761) Philosopher, theologian, mathematician Presbyterian (non-conformist) minister Tunbridge Wells, UK Elected FRS, perhaps due to
More informationBayesian Regression Linear and Logistic Regression
When we want more than point estimates Bayesian Regression Linear and Logistic Regression Nicole Beckage Ordinary Least Squares Regression and Lasso Regression return only point estimates But what if we
More informationContents. Part I: Fundamentals of Bayesian Inference 1
Contents Preface xiii Part I: Fundamentals of Bayesian Inference 1 1 Probability and inference 3 1.1 The three steps of Bayesian data analysis 3 1.2 General notation for statistical inference 4 1.3 Bayesian
More informationBayesian Methods for Machine Learning
Bayesian Methods for Machine Learning CS 584: Big Data Analytics Material adapted from Radford Neal s tutorial (http://ftp.cs.utoronto.ca/pub/radford/bayes-tut.pdf), Zoubin Ghahramni (http://hunch.net/~coms-4771/zoubin_ghahramani_bayesian_learning.pdf),
More informationPart 8: GLMs and Hierarchical LMs and GLMs
Part 8: GLMs and Hierarchical LMs and GLMs 1 Example: Song sparrow reproductive success Arcese et al., (1992) provide data on a sample from a population of 52 female song sparrows studied over the course
More informationBayesian Inference for DSGE Models. Lawrence J. Christiano
Bayesian Inference for DSGE Models Lawrence J. Christiano Outline State space-observer form. convenient for model estimation and many other things. Bayesian inference Bayes rule. Monte Carlo integation.
More informationComputational statistics
Computational statistics Markov Chain Monte Carlo methods Thierry Denœux March 2017 Thierry Denœux Computational statistics March 2017 1 / 71 Contents of this chapter When a target density f can be evaluated
More informationParameter Estimation. William H. Jefferys University of Texas at Austin Parameter Estimation 7/26/05 1
Parameter Estimation William H. Jefferys University of Texas at Austin bill@bayesrules.net Parameter Estimation 7/26/05 1 Elements of Inference Inference problems contain two indispensable elements: Data
More informationSTA 4273H: Sta-s-cal Machine Learning
STA 4273H: Sta-s-cal Machine Learning Russ Salakhutdinov Department of Computer Science! Department of Statistical Sciences! rsalakhu@cs.toronto.edu! h0p://www.cs.utoronto.ca/~rsalakhu/ Lecture 2 In our
More informationPhysics 403. Segev BenZvi. Numerical Methods, Maximum Likelihood, and Least Squares. Department of Physics and Astronomy University of Rochester
Physics 403 Numerical Methods, Maximum Likelihood, and Least Squares Segev BenZvi Department of Physics and Astronomy University of Rochester Table of Contents 1 Review of Last Class Quadratic Approximation
More informationDefault Priors and Effcient Posterior Computation in Bayesian
Default Priors and Effcient Posterior Computation in Bayesian Factor Analysis January 16, 2010 Presented by Eric Wang, Duke University Background and Motivation A Brief Review of Parameter Expansion Literature
More informationMarkov Chain Monte Carlo methods
Markov Chain Monte Carlo methods By Oleg Makhnin 1 Introduction a b c M = d e f g h i 0 f(x)dx 1.1 Motivation 1.1.1 Just here Supresses numbering 1.1.2 After this 1.2 Literature 2 Method 2.1 New math As
More informationStatistical Methods in Particle Physics Lecture 1: Bayesian methods
Statistical Methods in Particle Physics Lecture 1: Bayesian methods SUSSP65 St Andrews 16 29 August 2009 Glen Cowan Physics Department Royal Holloway, University of London g.cowan@rhul.ac.uk www.pp.rhul.ac.uk/~cowan
More informationeqr094: Hierarchical MCMC for Bayesian System Reliability
eqr094: Hierarchical MCMC for Bayesian System Reliability Alyson G. Wilson Statistical Sciences Group, Los Alamos National Laboratory P.O. Box 1663, MS F600 Los Alamos, NM 87545 USA Phone: 505-667-9167
More informationBAYESIAN METHODS FOR VARIABLE SELECTION WITH APPLICATIONS TO HIGH-DIMENSIONAL DATA
BAYESIAN METHODS FOR VARIABLE SELECTION WITH APPLICATIONS TO HIGH-DIMENSIONAL DATA Intro: Course Outline and Brief Intro to Marina Vannucci Rice University, USA PASI-CIMAT 04/28-30/2010 Marina Vannucci
More informationMarkov Chain Monte Carlo
Markov Chain Monte Carlo Recall: To compute the expectation E ( h(y ) ) we use the approximation E(h(Y )) 1 n n h(y ) t=1 with Y (1),..., Y (n) h(y). Thus our aim is to sample Y (1),..., Y (n) from f(y).
More informationMultivariate Bayes Wavelet Shrinkage and Applications
Journal of Applied Statistics Vol. 32, No. 5, 529 542, July 2005 Multivariate Bayes Wavelet Shrinkage and Applications GABRIEL HUERTA Department of Mathematics and Statistics, University of New Mexico
More informationMONTE CARLO METHODS. Hedibert Freitas Lopes
MONTE CARLO METHODS Hedibert Freitas Lopes The University of Chicago Booth School of Business 5807 South Woodlawn Avenue, Chicago, IL 60637 http://faculty.chicagobooth.edu/hedibert.lopes hlopes@chicagobooth.edu
More informationA quick introduction to Markov chains and Markov chain Monte Carlo (revised version)
A quick introduction to Markov chains and Markov chain Monte Carlo (revised version) Rasmus Waagepetersen Institute of Mathematical Sciences Aalborg University 1 Introduction These notes are intended to
More informationIntegrated Non-Factorized Variational Inference
Integrated Non-Factorized Variational Inference Shaobo Han, Xuejun Liao and Lawrence Carin Duke University February 27, 2014 S. Han et al. Integrated Non-Factorized Variational Inference February 27, 2014
More informationBayesian Inference for DSGE Models. Lawrence J. Christiano
Bayesian Inference for DSGE Models Lawrence J. Christiano Outline State space-observer form. convenient for model estimation and many other things. Preliminaries. Probabilities. Maximum Likelihood. Bayesian
More informationBayesian Inference and MCMC
Bayesian Inference and MCMC Aryan Arbabi Partly based on MCMC slides from CSC412 Fall 2018 1 / 18 Bayesian Inference - Motivation Consider we have a data set D = {x 1,..., x n }. E.g each x i can be the
More informationProbabilistic Graphical Networks: Definitions and Basic Results
This document gives a cursory overview of Probabilistic Graphical Networks. The material has been gleaned from different sources. I make no claim to original authorship of this material. Bayesian Graphical
More information(5) Multi-parameter models - Gibbs sampling. ST440/540: Applied Bayesian Analysis
Summarizing a posterior Given the data and prior the posterior is determined Summarizing the posterior gives parameter estimates, intervals, and hypothesis tests Most of these computations are integrals
More informationIntroduction to Bayesian Statistics and Markov Chain Monte Carlo Estimation. EPSY 905: Multivariate Analysis Spring 2016 Lecture #10: April 6, 2016
Introduction to Bayesian Statistics and Markov Chain Monte Carlo Estimation EPSY 905: Multivariate Analysis Spring 2016 Lecture #10: April 6, 2016 EPSY 905: Intro to Bayesian and MCMC Today s Class An
More informationBayesian Methods in Multilevel Regression
Bayesian Methods in Multilevel Regression Joop Hox MuLOG, 15 september 2000 mcmc What is Statistics?! Statistics is about uncertainty To err is human, to forgive divine, but to include errors in your design
More informationBayesian Phylogenetics:
Bayesian Phylogenetics: an introduction Marc A. Suchard msuchard@ucla.edu UCLA Who is this man? How sure are you? The one true tree? Methods we ve learned so far try to find a single tree that best describes
More informationMarkov Chain Monte Carlo in Practice
Markov Chain Monte Carlo in Practice Edited by W.R. Gilks Medical Research Council Biostatistics Unit Cambridge UK S. Richardson French National Institute for Health and Medical Research Vilejuif France
More informationBrief introduction to Markov Chain Monte Carlo
Brief introduction to Department of Probability and Mathematical Statistics seminar Stochastic modeling in economics and finance November 7, 2011 Brief introduction to Content 1 and motivation Classical
More informationBayesian Estimation with Sparse Grids
Bayesian Estimation with Sparse Grids Kenneth L. Judd and Thomas M. Mertens Institute on Computational Economics August 7, 27 / 48 Outline Introduction 2 Sparse grids Construction Integration with sparse
More informationLearning the hyper-parameters. Luca Martino
Learning the hyper-parameters Luca Martino 2017 2017 1 / 28 Parameters and hyper-parameters 1. All the described methods depend on some choice of hyper-parameters... 2. For instance, do you recall λ (bandwidth
More informationLikelihood-free MCMC
Bayesian inference for stable distributions with applications in finance Department of Mathematics University of Leicester September 2, 2011 MSc project final presentation Outline 1 2 3 4 Classical Monte
More informationLecture 5. G. Cowan Lectures on Statistical Data Analysis Lecture 5 page 1
Lecture 5 1 Probability (90 min.) Definition, Bayes theorem, probability densities and their properties, catalogue of pdfs, Monte Carlo 2 Statistical tests (90 min.) general concepts, test statistics,
More informationNonlinear Inequality Constrained Ridge Regression Estimator
The International Conference on Trends and Perspectives in Linear Statistical Inference (LinStat2014) 24 28 August 2014 Linköping, Sweden Nonlinear Inequality Constrained Ridge Regression Estimator Dr.
More informationBayesian inference for multivariate extreme value distributions
Bayesian inference for multivariate extreme value distributions Sebastian Engelke Clément Dombry, Marco Oesting Toronto, Fields Institute, May 4th, 2016 Main motivation For a parametric model Z F θ of
More informationMarkov Chain Monte Carlo
Markov Chain Monte Carlo (and Bayesian Mixture Models) David M. Blei Columbia University October 14, 2014 We have discussed probabilistic modeling, and have seen how the posterior distribution is the critical
More informationSupplement to A Hierarchical Approach for Fitting Curves to Response Time Measurements
Supplement to A Hierarchical Approach for Fitting Curves to Response Time Measurements Jeffrey N. Rouder Francis Tuerlinckx Paul L. Speckman Jun Lu & Pablo Gomez May 4 008 1 The Weibull regression model
More informationHastings-within-Gibbs Algorithm: Introduction and Application on Hierarchical Model
UNIVERSITY OF TEXAS AT SAN ANTONIO Hastings-within-Gibbs Algorithm: Introduction and Application on Hierarchical Model Liang Jing April 2010 1 1 ABSTRACT In this paper, common MCMC algorithms are introduced
More informationSTA 4273H: Statistical Machine Learning
STA 4273H: Statistical Machine Learning Russ Salakhutdinov Department of Computer Science! Department of Statistical Sciences! rsalakhu@cs.toronto.edu! h0p://www.cs.utoronto.ca/~rsalakhu/ Lecture 7 Approximate
More informationBayesian Inference for DSGE Models. Lawrence J. Christiano
Bayesian Inference for DSGE Models Lawrence J. Christiano Outline State space-observer form. convenient for model estimation and many other things. Bayesian inference Bayes rule. Monte Carlo integation.
More information7. Estimation and hypothesis testing. Objective. Recommended reading
7. Estimation and hypothesis testing Objective In this chapter, we show how the election of estimators can be represented as a decision problem. Secondly, we consider the problem of hypothesis testing
More informationSimulation of truncated normal variables. Christian P. Robert LSTA, Université Pierre et Marie Curie, Paris
Simulation of truncated normal variables Christian P. Robert LSTA, Université Pierre et Marie Curie, Paris Abstract arxiv:0907.4010v1 [stat.co] 23 Jul 2009 We provide in this paper simulation algorithms
More informationCSC 2541: Bayesian Methods for Machine Learning
CSC 2541: Bayesian Methods for Machine Learning Radford M. Neal, University of Toronto, 2011 Lecture 3 More Markov Chain Monte Carlo Methods The Metropolis algorithm isn t the only way to do MCMC. We ll
More informationMarkov Chain Monte Carlo (MCMC) and Model Evaluation. August 15, 2017
Markov Chain Monte Carlo (MCMC) and Model Evaluation August 15, 2017 Frequentist Linking Frequentist and Bayesian Statistics How can we estimate model parameters and what does it imply? Want to find the
More informationMarkov Networks.
Markov Networks www.biostat.wisc.edu/~dpage/cs760/ Goals for the lecture you should understand the following concepts Markov network syntax Markov network semantics Potential functions Partition function
More informationAccept-Reject Metropolis-Hastings Sampling and Marginal Likelihood Estimation
Accept-Reject Metropolis-Hastings Sampling and Marginal Likelihood Estimation Siddhartha Chib John M. Olin School of Business, Washington University, Campus Box 1133, 1 Brookings Drive, St. Louis, MO 63130.
More informationPrinciples of Bayesian Inference
Principles of Bayesian Inference Sudipto Banerjee 1 and Andrew O. Finley 2 1 Biostatistics, School of Public Health, University of Minnesota, Minneapolis, Minnesota, U.S.A. 2 Department of Forestry & Department
More informationPrinciples of Bayesian Inference
Principles of Bayesian Inference Sudipto Banerjee and Andrew O. Finley 2 Biostatistics, School of Public Health, University of Minnesota, Minneapolis, Minnesota, U.S.A. 2 Department of Forestry & Department
More informationBeyond MCMC in fitting complex Bayesian models: The INLA method
Beyond MCMC in fitting complex Bayesian models: The INLA method Valeska Andreozzi Centre of Statistics and Applications of Lisbon University (valeska.andreozzi at fc.ul.pt) European Congress of Epidemiology
More informationKernel adaptive Sequential Monte Carlo
Kernel adaptive Sequential Monte Carlo Ingmar Schuster (Paris Dauphine) Heiko Strathmann (University College London) Brooks Paige (Oxford) Dino Sejdinovic (Oxford) December 7, 2015 1 / 36 Section 1 Outline
More informationKobe University Repository : Kernel
Kobe University Repository : Kernel タイトル Title 著者 Author(s) 掲載誌 巻号 ページ Citation 刊行日 Issue date 資源タイプ Resource Type 版区分 Resource Version 権利 Rights DOI URL Note on the Sampling Distribution for the Metropolis-
More informationIntroduction to Machine Learning
Introduction to Machine Learning Brown University CSCI 1950-F, Spring 2012 Prof. Erik Sudderth Lecture 25: Markov Chain Monte Carlo (MCMC) Course Review and Advanced Topics Many figures courtesy Kevin
More informationBayesian data analysis in practice: Three simple examples
Bayesian data analysis in practice: Three simple examples Martin P. Tingley Introduction These notes cover three examples I presented at Climatea on 5 October 0. Matlab code is available by request to
More informationPattern Recognition and Machine Learning
Christopher M. Bishop Pattern Recognition and Machine Learning ÖSpri inger Contents Preface Mathematical notation Contents vii xi xiii 1 Introduction 1 1.1 Example: Polynomial Curve Fitting 4 1.2 Probability
More informationRank Regression with Normal Residuals using the Gibbs Sampler
Rank Regression with Normal Residuals using the Gibbs Sampler Stephen P Smith email: hucklebird@aol.com, 2018 Abstract Yu (2000) described the use of the Gibbs sampler to estimate regression parameters
More informationMarkov chain Monte Carlo methods in atmospheric remote sensing
1 / 45 Markov chain Monte Carlo methods in atmospheric remote sensing Johanna Tamminen johanna.tamminen@fmi.fi ESA Summer School on Earth System Monitoring and Modeling July 3 Aug 11, 212, Frascati July,
More informationApproximating Bayesian Posterior Means Using Multivariate Gaussian Quadrature
Approximating Bayesian Posterior Means Using Multivariate Gaussian Quadrature John A.L. Cranfield Paul V. Preckel Songquan Liu Presented at Western Agricultural Economics Association 1997 Annual Meeting
More informationCSC 2541: Bayesian Methods for Machine Learning
CSC 2541: Bayesian Methods for Machine Learning Radford M. Neal, University of Toronto, 2011 Lecture 10 Alternatives to Monte Carlo Computation Since about 1990, Markov chain Monte Carlo has been the dominant
More informationFundamental Issues in Bayesian Functional Data Analysis. Dennis D. Cox Rice University
Fundamental Issues in Bayesian Functional Data Analysis Dennis D. Cox Rice University 1 Introduction Question: What are functional data? Answer: Data that are functions of a continuous variable.... say
More informationThe Bayesian Approach to Multi-equation Econometric Model Estimation
Journal of Statistical and Econometric Methods, vol.3, no.1, 2014, 85-96 ISSN: 2241-0384 (print), 2241-0376 (online) Scienpress Ltd, 2014 The Bayesian Approach to Multi-equation Econometric Model Estimation
More informationSUPPLEMENT TO MARKET ENTRY COSTS, PRODUCER HETEROGENEITY, AND EXPORT DYNAMICS (Econometrica, Vol. 75, No. 3, May 2007, )
Econometrica Supplementary Material SUPPLEMENT TO MARKET ENTRY COSTS, PRODUCER HETEROGENEITY, AND EXPORT DYNAMICS (Econometrica, Vol. 75, No. 3, May 2007, 653 710) BY SANGHAMITRA DAS, MARK ROBERTS, AND
More informationMarkov Chain Monte Carlo
Chapter 5 Markov Chain Monte Carlo MCMC is a kind of improvement of the Monte Carlo method By sampling from a Markov chain whose stationary distribution is the desired sampling distributuion, it is possible
More informationEstimating marginal likelihoods from the posterior draws through a geometric identity
Estimating marginal likelihoods from the posterior draws through a geometric identity Johannes Reichl Energy Institute at the Johannes Kepler University Linz E-mail for correspondence: reichl@energieinstitut-linz.at
More informationBayesian Inference in GLMs. Frequentists typically base inferences on MLEs, asymptotic confidence
Bayesian Inference in GLMs Frequentists typically base inferences on MLEs, asymptotic confidence limits, and log-likelihood ratio tests Bayesians base inferences on the posterior distribution of the unknowns
More informationSystematic uncertainties in statistical data analysis for particle physics. DESY Seminar Hamburg, 31 March, 2009
Systematic uncertainties in statistical data analysis for particle physics DESY Seminar Hamburg, 31 March, 2009 Glen Cowan Physics Department Royal Holloway, University of London g.cowan@rhul.ac.uk www.pp.rhul.ac.uk/~cowan
More informationSampling Methods (11/30/04)
CS281A/Stat241A: Statistical Learning Theory Sampling Methods (11/30/04) Lecturer: Michael I. Jordan Scribe: Jaspal S. Sandhu 1 Gibbs Sampling Figure 1: Undirected and directed graphs, respectively, with
More informationPart 1: Expectation Propagation
Chalmers Machine Learning Summer School Approximate message passing and biomedicine Part 1: Expectation Propagation Tom Heskes Machine Learning Group, Institute for Computing and Information Sciences Radboud
More informationHierarchical Bayesian approaches for robust inference in ARX models
Hierarchical Bayesian approaches for robust inference in ARX models Johan Dahlin, Fredrik Lindsten, Thomas Bo Schön and Adrian George Wills Linköping University Post Print N.B.: When citing this work,
More informationMetropolis-Hastings Algorithm
Strength of the Gibbs sampler Metropolis-Hastings Algorithm Easy algorithm to think about. Exploits the factorization properties of the joint probability distribution. No difficult choices to be made to
More informationA Bayesian Approach to Phylogenetics
A Bayesian Approach to Phylogenetics Niklas Wahlberg Based largely on slides by Paul Lewis (www.eeb.uconn.edu) An Introduction to Bayesian Phylogenetics Bayesian inference in general Markov chain Monte
More informationSTA 4273H: Statistical Machine Learning
STA 4273H: Statistical Machine Learning Russ Salakhutdinov Department of Statistics! rsalakhu@utstat.toronto.edu! http://www.utstat.utoronto.ca/~rsalakhu/ Sidney Smith Hall, Room 6002 Lecture 3 Linear
More informationMCMC algorithms for fitting Bayesian models
MCMC algorithms for fitting Bayesian models p. 1/1 MCMC algorithms for fitting Bayesian models Sudipto Banerjee sudiptob@biostat.umn.edu University of Minnesota MCMC algorithms for fitting Bayesian models
More informationBayesian Nonparametric Regression for Diabetes Deaths
Bayesian Nonparametric Regression for Diabetes Deaths Brian M. Hartman PhD Student, 2010 Texas A&M University College Station, TX, USA David B. Dahl Assistant Professor Texas A&M University College Station,
More informationBayesian linear regression
Bayesian linear regression Linear regression is the basis of most statistical modeling. The model is Y i = X T i β + ε i, where Y i is the continuous response X i = (X i1,..., X ip ) T is the corresponding
More informationThe STS Surgeon Composite Technical Appendix
The STS Surgeon Composite Technical Appendix Overview Surgeon-specific risk-adjusted operative operative mortality and major complication rates were estimated using a bivariate random-effects logistic
More informationLecture 6: Markov Chain Monte Carlo
Lecture 6: Markov Chain Monte Carlo D. Jason Koskinen koskinen@nbi.ku.dk Photo by Howard Jackman University of Copenhagen Advanced Methods in Applied Statistics Feb - Apr 2016 Niels Bohr Institute 2 Outline
More informationDown by the Bayes, where the Watermelons Grow
Down by the Bayes, where the Watermelons Grow A Bayesian example using SAS SUAVe: Victoria SAS User Group Meeting November 21, 2017 Peter K. Ott, M.Sc., P.Stat. Strategic Analysis 1 Outline 1. Motivating
More informationEstimating the marginal likelihood with Integrated nested Laplace approximation (INLA)
Estimating the marginal likelihood with Integrated nested Laplace approximation (INLA) arxiv:1611.01450v1 [stat.co] 4 Nov 2016 Aliaksandr Hubin Department of Mathematics, University of Oslo and Geir Storvik
More information6.1 Approximating the posterior by calculating on a grid
Chapter 6 Approximations 6.1 Approximating the posterior by calculating on a grid When one is confronted with a low-dimensional problem with a continuous parameter, then it is usually feasible to approximate
More informationMarkov chain Monte Carlo methods for visual tracking
Markov chain Monte Carlo methods for visual tracking Ray Luo rluo@cory.eecs.berkeley.edu Department of Electrical Engineering and Computer Sciences University of California, Berkeley Berkeley, CA 94720
More informationAn introduction to Bayesian statistics and model calibration and a host of related topics
An introduction to Bayesian statistics and model calibration and a host of related topics Derek Bingham Statistics and Actuarial Science Simon Fraser University Cast of thousands have participated in the
More informationσ(a) = a N (x; 0, 1 2 ) dx. σ(a) = Φ(a) =
Until now we have always worked with likelihoods and prior distributions that were conjugate to each other, allowing the computation of the posterior distribution to be done in closed form. Unfortunately,
More informationNovember 2002 STA Random Effects Selection in Linear Mixed Models
November 2002 STA216 1 Random Effects Selection in Linear Mixed Models November 2002 STA216 2 Introduction It is common practice in many applications to collect multiple measurements on a subject. Linear
More informationCPSC 540: Machine Learning
CPSC 540: Machine Learning MCMC and Non-Parametric Bayes Mark Schmidt University of British Columbia Winter 2016 Admin I went through project proposals: Some of you got a message on Piazza. No news is
More informationThe Bayesian Choice. Christian P. Robert. From Decision-Theoretic Foundations to Computational Implementation. Second Edition.
Christian P. Robert The Bayesian Choice From Decision-Theoretic Foundations to Computational Implementation Second Edition With 23 Illustrations ^Springer" Contents Preface to the Second Edition Preface
More informationNonparametric Bayesian Methods (Gaussian Processes)
[70240413 Statistical Machine Learning, Spring, 2015] Nonparametric Bayesian Methods (Gaussian Processes) Jun Zhu dcszj@mail.tsinghua.edu.cn http://bigml.cs.tsinghua.edu.cn/~jun State Key Lab of Intelligent
More informationKatsuhiro Sugita Faculty of Law and Letters, University of the Ryukyus. Abstract
Bayesian analysis of a vector autoregressive model with multiple structural breaks Katsuhiro Sugita Faculty of Law and Letters, University of the Ryukyus Abstract This paper develops a Bayesian approach
More informationBayesian System Identification based on Hierarchical Sparse Bayesian Learning and Gibbs Sampling with Application to Structural Damage Assessment
Bayesian System Identification based on Hierarchical Sparse Bayesian Learning and Gibbs Sampling with Application to Structural Damage Assessment Yong Huang a,b, James L. Beck b,* and Hui Li a a Key Lab
More informationUnivariate Normal Distribution; GLM with the Univariate Normal; Least Squares Estimation
Univariate Normal Distribution; GLM with the Univariate Normal; Least Squares Estimation PRE 905: Multivariate Analysis Spring 2014 Lecture 4 Today s Class The building blocks: The basics of mathematical
More informationML estimation: Random-intercepts logistic model. and z
ML estimation: Random-intercepts logistic model log p ij 1 p = x ijβ + υ i with υ i N(0, συ) 2 ij Standardizing the random effect, θ i = υ i /σ υ, yields log p ij 1 p = x ij β + σ υθ i with θ i N(0, 1)
More informationWho was Bayes? Bayesian Phylogenetics. What is Bayes Theorem?
Who was Bayes? Bayesian Phylogenetics Bret Larget Departments of Botany and of Statistics University of Wisconsin Madison October 6, 2011 The Reverand Thomas Bayes was born in London in 1702. He was the
More informationGeneralized Linear Models for Non-Normal Data
Generalized Linear Models for Non-Normal Data Today s Class: 3 parts of a generalized model Models for binary outcomes Complications for generalized multivariate or multilevel models SPLH 861: Lecture
More information