Bayesian Meta-analysis with Hierarchical Modeling Brian P. Hobbs 1

Size: px
Start display at page:

Download "Bayesian Meta-analysis with Hierarchical Modeling Brian P. Hobbs 1"

Transcription

1 Bayesian Meta-analysis with Hierarchical Modeling Brian P. Hobbs 1 Division of Biostatistics, School of Public Health, University of Minnesota, Mayo Mail Code 303, Minneapolis, Minnesota , U.S.A. 1 Brian P. Hobbs is Graduate Assistant and Bradley P. Carlin is Professor of Biostatistics and Mayo Professor in Public Health at the Division of Biostatistics, School of Public Health, 420 Delaware St. S.E., University of Minnesota, Minneapolis, MN,

2 1 Introduction The Bayesian approach to inference enables relevant existing information to be formally incorporated into a statistical analysis. This is done through the specification of prior distributions, which summarize our preexisting understanding or beliefs regarding any unknown model parameters θ = (θ 1,..., θ K ). Inference is conducted on the posterior distribution of θ given the observed data y = (y 1,..., y N ), given by Bayes Rule as p(θ y) = p(θ, y) p(y) = p(y θ)p(θ). p(y θ)p(θ) This simple formulation assumes the prior p(θ) is fully specified. However, when we are less certain about p(θ), or when model variability must be allocated to multiple sources (say, centers and patients within centers), a hierarchical model may be more appropriate. This approach places prior distributions on the unknown parameters of previously specified priors in stages. Posterior distributions are again derived by Bayes theorem, where the denominator integral is now more difficult, but remains feasible using modern Markov chain Monte Carlo (MCMC) methods. The WinBUGS package ( and its open source cousin OpenBUGS ( are able to handle a wide variety of hierarchical models, permitting posterior inference, prediction, model choice, and model checking all within a user-friendly MCMC framework. Hierarchical models permit borrowing of strength from the prior distributions and across subgroups. When combined with any information incorporated in the priors, this translates into a larger effective sample size, thus offering potentially important savings (both ethical and financial) in the practice of drug and device clinical trials. Suppose the prior distribution for θ depends on a vector of second-stage parameters γ. These parameters are called hyperparameters, and we then write p(θ γ). In a simple two-stage model, γ is assumed to be known, and is often set to produce a noninformative prior (i.e., one that does not favor one value of θ over any other. However, if γ is unknown, a third-stage prior, or hyperprior, p(γ) may be chosen. In clinical trials, p(γ) is often determined at least in part using data from existing historical controls. This additional informative content is part of what gives Bayesian methods their advantage over classical methods, although this advantage is typically small if noninformative priors are used. 1

3 2 Application of Bayes methods to Meta-analysis Results vary across studies due to random variation or differences in implementation. The studies may be carried out at different times and locations or include different types of subjects. Furthermore, the applications of eligibility criteria may vary. These differences may lead to disparate conclusions about the intervention of interest across studies. Consider, for example two studies which test the ability of a particular cardiac device to improve heart efficiency by increasing the amount of blood pumped out of the left ventricle relative to the amount blood contained in the ventricle. Suppose that both studies define eligible patients as those who have a left ventricular ejection fraction (LVEF) as low as 25%. One investigator may admit every such eligible candidate patient. A second investigator might alter the LVEF boundary to 40% for a subset of individuals with another condition, restricting eligibility. Consequently, the first study may tend to incorporate a frailer population, so that the first study may suggest that the device is less effective than the second (Berry, 1997). In such cases, a single comprehensive analysis of all relevant data from several independent studies, or a meta-analysis, is often used to assess the clinical effectiveness of healthcare interventions. Results from metaanalyses provide a thorough assessment of the intervention of interest. The Bayesian hierarchical approach to meta-analysis treats study as one level of experimental unit, and patient within study as a second level (Lindley and Smith, 1972; Berger, 1985). Inter-study differences may be accounted for by measured covariates as in the above illustration; however, unaccounted for differences will still remain. Since the Bayesian paradigm treats all unknowns as random, a Bayesian meta-analysis can be structured as a random effects model. Specifically, each study in a Bayesian meta-analysis has a distribution of patient responses specific to the particular study. Thus selecting a study corresponds to selecting one of these distributions. Furthermore, one is limited to only a sample from each study s distribution, revealing only indirect information about the distribution of study-specific effects. 2.1 Meta-analysis for a Single Success Proportion Berry (1997, Sec. 3.1) describes a simple yet commonly occurring setting where Bayesian meta-analysis pays significant dividends, illustrating with the data in Table 1. These data are from nine antidepressant drug 2

4 Study (i) x i n i ˆπ i = x i /n i Total Table 1: Successes x i and total numbers of patients n i in 9 antidepressant drug studies. studies (Janicak et al., 1988), where a success is considered a positive response to the treatment regimen. For our purpose of illustrating Bayesian hierarchical modeling in meta-analysis, suppose a success concerns effectiveness of a medical device, and that within study i the experimental units receiving the intervention are exchangeable (all have the same probability of success π i ). Define the random variable x i to be the number of successes among the n i patients in study i, so that x i Binomial(n i, π i ) for i = 1,..., 9. The likelihood function for π = (π 1,..., π n ) is then p(x π) n i=1 π x i i (1 π i ) n i x i (1) A pooled analysis assumes that all 150 patients are independent and identically distributed (iid). Therefore, all nine π i s are equal to a common π. Given we have 106 total successes in 150 trials, the likelihood function is then p(x π) π 106 (1 π) 44, and suggests that π is very likely to be between 0.6 and 0.8. However, the observed success proportions (Table 1) in five of the nine studies are outside this range. This is more than what would be expected from sampling variability alone, and suggests the π i s may be unequal. Sadly, separate analyses of the nine studies provides even less satisfying results. The effect of an experimental device is not well addressed by giving nine different likelihood functions, or by giving nine different 3

5 Figure 1: Beta(α, β) densities for α, β = 1, 2, 4, 8. confidence intervals. Consider the probability of success if the device were used in a tenth study with another patient population. Separate analyses provide no way to utilize the results from the nine previous studies. A Bayesian hierarchical perspective provides a beneficial middle ground. Here we view each study s success probability π i as having been selected from a population. A computationally convenient assumption here is to suppose the π i are random sample from a beta distribution, i.e., π i iid Beta(α, β). Denoting the beta function as B(α, β) = Γ(α+β)/[Γ(α)Γ(β)], for each π i, p(π i α, β) = B(α, β)πi α 1 (1 π i ) β 1, a beta distribution with mean E(π i α, β) = α α+β and variance V ar(π i α, β) = αβ (α+β) 2 (α+β+1), where α, β > 0. Since lim α+β [V ar(π i )] = 0, we can think of α + β as measuring homogeneity among studies. If α + β is large then the π i s distribution is highly concentrated near its mean. Smaller α and β permit more variability, hence a noticeable study effect (unequal π i ). Assuming that only two parameters index the entire distribution may be a restriction depending on the curve of choice. In this case, Figure 1 shows the beta family to be surprisingly flexible, able to capture various shapes (flat, bell-shaped, U-shaped, one-tailed, etc.). Since the Beta prior is conjugate with the binomial likelihood, the posterior of π i given x i emerges in closed form using Bayes theorem as p(π i x i ) π α 1+x i i (1 π i ) β 1+n i x i. That is, the π i are independent 4

6 Beta(α + x i, β + n i x i ) random variables with mean E πi [π i x i ] = α + x i α + β + n i. (2) In order to proceed further with a Bayesian hierarchical approach, the impact of the hyperprior p(α, β), for the second stage parameters needs to be assessed. Recall that concentrating the hyperprior s probability on large values of α and β suggests homogeneity among the π i. while small α + β suggests heterogeneity. If each π i was observable, then posterior distribution of (α, β) would be p(α, β π i ) 9 i=1 { B(α, β)π α 1 i (1 π i ) β 1} p(α, β). In reality, π i cannot be observed directly, but indirect information about π 1,..., π 9 is available through observations x = (x 1,..., x 9 ). Therefore, the posterior distribution of α and β, p(α, β x), is proportional to p(α, β) 9 B(α, β) i=1 1 0 π x i+α 1 i (1 π i ) n i x i +β 1 dπ 1 dπ 9 9 { i=1 B(α, β) B(α + x i, β + n i x i ) } p(α, β). (3) Given the data in Table 1, the posterior expected mean success rate for the next patient treated in study i, for i = 1,..., 9 is [ ] α + xi E(π i x) = E (α,β) {E πi [π i α, β, x]} = E (α,β) x, i = 1,..., 9. (4) α + β + n i Next, predictive distributions are obtained by averaging the likelihood over the full posterior distribution. If a new, tenth study similar to the first nine is implemented, inference for π 10 requires the predictive distribution, p(π 10 x 1,..., x 9 ) = p(π 10 α, β, x 1,..., x 9 )p(α, β x 1,..., x 9 )dαdβ. It follows that the expected probability of a successful treatment for a particular patient enrolled in the new study is the expected posterior mean of the predictive distribution of π 10, [ ] α E(π 10 x) = E (α,β) {E[π 10 α, β, x]} = E (α,β) α + β x. (5) 5

7 Hyperprior probability distributions are often hard to conceptualize. Commonly reference priors are used when assign distributions beyond the second stage. In the current model, the shape of the first stage prior distribution varies considerably for relatively small changes in α, β as seen in Figure 1. Therefore, a prior that associates some probability with α+β large and α+β small, while assigning a moderate amount of probability to roughly equivalent α and β will be quite effective in covering a wide range of shapes for p(π i α, β). Bowing to computational limitations of the time, Berry (1997) adopted independent discrete uniform priors on {1, 2,..., 10} for α and β, essentially discretizing the (α, β) space onto a square grid. Here we switch to independent continuous U(0, 20) priors, a true joint flat prior over a broad range of sensible values. Thus the posterior probability density function p(α, β x) is proportional to the likelihood restricted to [0, 20] [0, 20]. Note also that values larger than 20 have some likelihood, therefore the truncation of α and β at 20 is a slight approximation made by this model. 2.2 Sampling Based Inference using MCMC In order to analyze the data in Tables 1 using our three-stage model, we use Markov chain Monte Carlo (MCMC) computational methods implemented in WinBUGS. These methods operate by sampling from a Markov chain whose stationary distribution is the joint posterior distribution. This permits easy evaluation of posterior distributions for the π i, which lack closed forms due to the nonconjugate hyperprior for (α, β). Specifically, we may estimate E(π x x) in (4) by MCMC sampling {(α (g), β (g) ), g = 1,..., G} from their joint posterior, and then using the Monte Carlo approximation Ê(π i x) = 1 G G g=1 α (g) + x i α (g) + β (g) + n i. The Gibbs sampler begins the Markov chain with initial values ( inits ) (π (0), α (0), β (0) ), and then successively samples from the conditional distributions for α, β, and the π i. We usually discard draws from the first K iterations, the initial transient or burn-in period, though choosing reasonable initial values can reduce the need for this. Typically, multiple Markov chains are started from disparate initial values and checked to see if they all appear to have the same equilibrium distribution. Modern software packages make MCMC 6

8 Figure 2: Posterior summary statistics generated in WinBUGS given the data in Table 1. β α Figure 3: Bivariate posterior scatterplot of (α, β) given the data in Table 1. Correlation(α, β) = sampling quick and relatively easy; the popular WinBUGS package also offers several convergence checks and output summary tools. Using the WinBUGS code, data, and inits shown in Appendix A, we ran the Gibbs sampler for 30,000 iterations, discarding the first 10,000 as burn-in. Notice that we added a tenth study with no observations for n10 = 10 patients. Summary statistics for the posterior distributions of model parameters are shown in Figure 2. The posterior mean of p(α, β x) is (9.50, 4.30). Figure 3 shows that (α, β) given the data in Table 1 are highly correlated. The expected posterior predictive mean success rate for a patient in a new study, i = 10, is the posterior mean of τ = α/α + β, or posterior mean of π10, which from Figure 2 is approximately The posterior mean success rates (4) for each of the nine studies are also given in Figure 2. These posterior 7

9 means represent the probability that the intervention is successful for the next patient in the respective study. The and posterior quantiles are also given in Figure 2, permitting ready evaluation of equal-tail 95% Bayesian credible intervals for the π i. Regression (or shrinkage) to the overall mean is observed in the predictive probabilities for each of the nine studies. More shrinkage occurs for smaller studies, since their likelihoods are less informative; see e.g. the high shrinkage and relatively wide credible interval for study 2.!" #$π10 π Figure 4: Posterior density of π 10 (the next trial) given the data in Table 1. Finally, Figure 4 plots the posterior distribution of expected success proportion in the next study, i = 10, given the results of the previous nine studies, (5), as well as the likelihood function that assumes all nine studies have a common success probability π. The posterior distribution of π 10 clearly has more variability, and appears to be more consistent with the observed success proportions observed in Table 1. 8

10 References Berger, J.O. (1985). Statistical Decision Theory and Bayesian Analysis, 2nd ed. New York: Springer-Verlag. Berry, D.A. (1997). Using a Bayesian approach in medical device development. White paper, Center for Devices and Radiological Health, U.S. Food and Drug Administration, Rockville, MD. Berry, D.A. (2006). Bayesian clinical trials. Nature Reviews Drug Discovery, 5, Janicak, P.G., Lipinski, J., Davis, J.M., Comaty, J.E., Waternaux, C., Cohen, B., Altman, E., and Sharma, R.P. (1988). S-adenosyl-methionine (SAMe) in depression: a literature review and preliminary data report. Alabama Journal of Medical Sciences, 25, Lindley, D.V., and Smith, A.F.M. (1972). Bayes estimates for the linear model (with discussion). J. Roy. Statist. Soc., Ser. B, 34, Spiegelhalter, D.J., Abrams, K.R., and Myles, J.P. (2004). Bayesian Approaches to Clinical Trials and Health-Care Evaluation. Chichester: Wiley. Spiegelhalter, D.J., Best, N., Carlin, B.P., and van der Linde, A. (2002). Bayesian measures of model complexity and fit (with discussion). J. Roy. Statist. Soc., Ser. B, 64, Spiegelhalter, D.J., Freedman, L.S., and Parmar, M.K.B. (1994). Bayesian approaches to randomized trials (with discussion). J. Roy. Statist. Soc., Ser. A, 157,

11 Appendix A: WinBUGS code for the Meta-analysis example!" " # $%&''( ) ) *+,- #."/0" /" " /"12 5 4$

Principles of Bayesian Inference

Principles 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 information

Principles of Bayesian Inference

Principles 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 information

Principles of Bayesian Inference

Principles 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 information

eqr094: Hierarchical MCMC for Bayesian System Reliability

eqr094: 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 information

Principles of Bayesian Inference

Principles 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 information

Inference for a Population Proportion

Inference for a Population Proportion Al Nosedal. University of Toronto. November 11, 2015 Statistical inference is drawing conclusions about an entire population based on data in a sample drawn from that population. From both frequentist

More information

Bayesian Linear Regression

Bayesian Linear Regression Bayesian Linear Regression Sudipto Banerjee 1 Biostatistics, School of Public Health, University of Minnesota, Minneapolis, Minnesota, U.S.A. September 15, 2010 1 Linear regression models: a Bayesian perspective

More information

10. Exchangeability and hierarchical models Objective. Recommended reading

10. Exchangeability and hierarchical models Objective. Recommended reading 10. Exchangeability and hierarchical models Objective Introduce exchangeability and its relation to Bayesian hierarchical models. Show how to fit such models using fully and empirical Bayesian methods.

More information

MCMC algorithms for fitting Bayesian models

MCMC 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 information

Markov chain Monte Carlo

Markov chain Monte Carlo 1 / 26 Markov chain Monte Carlo Timothy Hanson 1 and Alejandro Jara 2 1 Division of Biostatistics, University of Minnesota, USA 2 Department of Statistics, Universidad de Concepción, Chile IAP-Workshop

More information

Markov Chain Monte Carlo (MCMC) and Model Evaluation. August 15, 2017

Markov 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 information

Pubh 8482: Sequential Analysis

Pubh 8482: Sequential Analysis Pubh 8482: Sequential Analysis Joseph S. Koopmeiners Division of Biostatistics University of Minnesota Week 10 Class Summary Last time... We began our discussion of adaptive clinical trials Specifically,

More information

Hierarchical Models & Bayesian Model Selection

Hierarchical Models & Bayesian Model Selection Hierarchical Models & Bayesian Model Selection Geoffrey Roeder Departments of Computer Science and Statistics University of British Columbia Jan. 20, 2016 Contact information Please report any typos or

More information

Bayesian methods for sample size determination and their use in clinical trials

Bayesian methods for sample size determination and their use in clinical trials Bayesian methods for sample size determination and their use in clinical trials Stefania Gubbiotti Abstract This paper deals with determination of a sample size that guarantees the success of a trial.

More information

POSTERIOR PROPRIETY IN SOME HIERARCHICAL EXPONENTIAL FAMILY MODELS

POSTERIOR PROPRIETY IN SOME HIERARCHICAL EXPONENTIAL FAMILY MODELS POSTERIOR PROPRIETY IN SOME HIERARCHICAL EXPONENTIAL FAMILY MODELS EDWARD I. GEORGE and ZUOSHUN ZHANG The University of Texas at Austin and Quintiles Inc. June 2 SUMMARY For Bayesian analysis of hierarchical

More information

Part 7: Hierarchical Modeling

Part 7: Hierarchical Modeling Part 7: Hierarchical Modeling!1 Nested data It is common for data to be nested: i.e., observations on subjects are organized by a hierarchy Such data are often called hierarchical or multilevel For example,

More information

Markov Chain Monte Carlo methods

Markov 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 information

Bayesian Methods for Machine Learning

Bayesian 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 information

7. Estimation and hypothesis testing. Objective. Recommended reading

7. 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 information

Bayesian philosophy Bayesian computation Bayesian software. Bayesian Statistics. Petter Mostad. Chalmers. April 6, 2017

Bayesian philosophy Bayesian computation Bayesian software. Bayesian Statistics. Petter Mostad. Chalmers. April 6, 2017 Chalmers April 6, 2017 Bayesian philosophy Bayesian philosophy Bayesian statistics versus classical statistics: War or co-existence? Classical statistics: Models have variables and parameters; these are

More information

Markov Chain Monte Carlo

Markov 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 information

Introduction to Bayesian Methods

Introduction to Bayesian Methods Introduction to Bayesian Methods Jessi Cisewski Department of Statistics Yale University Sagan Summer Workshop 2016 Our goal: introduction to Bayesian methods Likelihoods Priors: conjugate priors, non-informative

More information

Markov Chain Monte Carlo methods

Markov 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 information

A note on Reversible Jump Markov Chain Monte Carlo

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 information

θ 1 θ 2 θ n y i1 y i2 y in Hierarchical models (chapter 5) Hierarchical model Introduction to hierarchical models - sometimes called multilevel model

θ 1 θ 2 θ n y i1 y i2 y in Hierarchical models (chapter 5) Hierarchical model Introduction to hierarchical models - sometimes called multilevel model Hierarchical models (chapter 5) Introduction to hierarchical models - sometimes called multilevel model Exchangeability Slide 1 Hierarchical model Example: heart surgery in hospitals - in hospital j survival

More information

(5) Multi-parameter models - Gibbs sampling. ST440/540: Applied Bayesian Analysis

(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 information

Bayesian Estimation An Informal Introduction

Bayesian Estimation An Informal Introduction Mary Parker, Bayesian Estimation An Informal Introduction page 1 of 8 Bayesian Estimation An Informal Introduction Example: I take a coin out of my pocket and I want to estimate the probability of heads

More information

Hastings-within-Gibbs Algorithm: Introduction and Application on Hierarchical Model

Hastings-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 information

Bayesian Linear Models

Bayesian Linear Models Bayesian Linear Models Sudipto Banerjee 1 and Andrew O. Finley 2 1 Department of Forestry & Department of Geography, Michigan State University, Lansing Michigan, U.S.A. 2 Biostatistics, School of Public

More information

Contents. Part I: Fundamentals of Bayesian Inference 1

Contents. 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 information

Lecture 13 Fundamentals of Bayesian Inference

Lecture 13 Fundamentals of Bayesian Inference Lecture 13 Fundamentals of Bayesian Inference Dennis Sun Stats 253 August 11, 2014 Outline of Lecture 1 Bayesian Models 2 Modeling Correlations Using Bayes 3 The Universal Algorithm 4 BUGS 5 Wrapping Up

More information

Introduction to Bayesian Methods. Introduction to Bayesian Methods p.1/??

Introduction to Bayesian Methods. Introduction to Bayesian Methods p.1/?? to Bayesian Methods Introduction to Bayesian Methods p.1/?? We develop the Bayesian paradigm for parametric inference. To this end, suppose we conduct (or wish to design) a study, in which the parameter

More information

Bayesian Inference. Chapter 4: Regression and Hierarchical Models

Bayesian Inference. Chapter 4: Regression and Hierarchical Models Bayesian Inference Chapter 4: Regression and Hierarchical Models Conchi Ausín and Mike Wiper Department of Statistics Universidad Carlos III de Madrid Master in Business Administration and Quantitative

More information

Bayesian Linear Models

Bayesian Linear Models Bayesian Linear Models 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 information

Lecture 2: Priors and Conjugacy

Lecture 2: Priors and Conjugacy Lecture 2: Priors and Conjugacy Melih Kandemir melih.kandemir@iwr.uni-heidelberg.de May 6, 2014 Some nice courses Fred A. Hamprecht (Heidelberg U.) https://www.youtube.com/watch?v=j66rrnzzkow Michael I.

More information

Bayesian Inference. Chapter 1. Introduction and basic concepts

Bayesian Inference. Chapter 1. Introduction and basic concepts Bayesian Inference Chapter 1. Introduction and basic concepts M. Concepción Ausín Department of Statistics Universidad Carlos III de Madrid Master in Business Administration and Quantitative Methods Master

More information

Introduction: MLE, MAP, Bayesian reasoning (28/8/13)

Introduction: MLE, MAP, Bayesian reasoning (28/8/13) STA561: Probabilistic machine learning Introduction: MLE, MAP, Bayesian reasoning (28/8/13) Lecturer: Barbara Engelhardt Scribes: K. Ulrich, J. Subramanian, N. Raval, J. O Hollaren 1 Classifiers In this

More information

PARAMETER ESTIMATION: BAYESIAN APPROACH. These notes summarize the lectures on Bayesian parameter estimation.

PARAMETER ESTIMATION: BAYESIAN APPROACH. These notes summarize the lectures on Bayesian parameter estimation. PARAMETER ESTIMATION: BAYESIAN APPROACH. These notes summarize the lectures on Bayesian parameter estimation.. Beta Distribution We ll start by learning about the Beta distribution, since we end up using

More information

Bayesian Inference. Chapter 4: Regression and Hierarchical Models

Bayesian Inference. Chapter 4: Regression and Hierarchical Models Bayesian Inference Chapter 4: Regression and Hierarchical Models Conchi Ausín and Mike Wiper Department of Statistics Universidad Carlos III de Madrid Advanced Statistics and Data Mining Summer School

More information

The Mixture Approach for Simulating New Families of Bivariate Distributions with Specified Correlations

The Mixture Approach for Simulating New Families of Bivariate Distributions with Specified Correlations The Mixture Approach for Simulating New Families of Bivariate Distributions with Specified Correlations John R. Michael, Significance, Inc. and William R. Schucany, Southern Methodist University The mixture

More information

McGill University. Department of Epidemiology and Biostatistics. Bayesian Analysis for the Health Sciences. Course EPIB-675.

McGill University. Department of Epidemiology and Biostatistics. Bayesian Analysis for the Health Sciences. Course EPIB-675. McGill University Department of Epidemiology and Biostatistics Bayesian Analysis for the Health Sciences Course EPIB-675 Lawrence Joseph Bayesian Analysis for the Health Sciences EPIB-675 3 credits Instructor:

More information

Chapter 5. Bayesian Statistics

Chapter 5. Bayesian Statistics Chapter 5. Bayesian Statistics Principles of Bayesian Statistics Anything unknown is given a probability distribution, representing degrees of belief [subjective probability]. Degrees of belief [subjective

More information

Reminder of some Markov Chain properties:

Reminder of some Markov Chain properties: Reminder of some Markov Chain properties: 1. a transition from one state to another occurs probabilistically 2. only state that matters is where you currently are (i.e. given present, future is independent

More information

Bayesian Networks in Educational Assessment

Bayesian Networks in Educational Assessment Bayesian Networks in Educational Assessment Estimating Parameters with MCMC Bayesian Inference: Expanding Our Context Roy Levy Arizona State University Roy.Levy@asu.edu 2017 Roy Levy MCMC 1 MCMC 2 Posterior

More information

36-463/663Multilevel and Hierarchical Models

36-463/663Multilevel and Hierarchical Models 36-463/663Multilevel and Hierarchical Models From Bayes to MCMC to MLMs Brian Junker 132E Baker Hall brian@stat.cmu.edu 1 Outline Bayesian Statistics and MCMC Distribution of Skill Mastery in a Population

More information

The Bayesian Approach to Multi-equation Econometric Model Estimation

The 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 information

2 Bayesian Hierarchical Response Modeling

2 Bayesian Hierarchical Response Modeling 2 Bayesian Hierarchical Response Modeling In the first chapter, an introduction to Bayesian item response modeling was given. The Bayesian methodology requires careful specification of priors since item

More information

Bayesian Phylogenetics:

Bayesian 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 information

Computational statistics

Computational 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 information

Integrated Non-Factorized Variational Inference

Integrated 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 information

David Giles Bayesian Econometrics

David Giles Bayesian Econometrics David Giles Bayesian Econometrics 1. General Background 2. Constructing Prior Distributions 3. Properties of Bayes Estimators and Tests 4. Bayesian Analysis of the Multiple Regression Model 5. Bayesian

More information

Using Model Selection and Prior Specification to Improve Regime-switching Asset Simulations

Using Model Selection and Prior Specification to Improve Regime-switching Asset Simulations Using Model Selection and Prior Specification to Improve Regime-switching Asset Simulations Brian M. Hartman, PhD ASA Assistant Professor of Actuarial Science University of Connecticut BYU Statistics Department

More information

Ronald Christensen. University of New Mexico. Albuquerque, New Mexico. Wesley Johnson. University of California, Irvine. Irvine, California

Ronald Christensen. University of New Mexico. Albuquerque, New Mexico. Wesley Johnson. University of California, Irvine. Irvine, California Texts in Statistical Science Bayesian Ideas and Data Analysis An Introduction for Scientists and Statisticians Ronald Christensen University of New Mexico Albuquerque, New Mexico Wesley Johnson University

More information

Hyperparameter estimation in Dirichlet process mixture models

Hyperparameter estimation in Dirichlet process mixture models Hyperparameter estimation in Dirichlet process mixture models By MIKE WEST Institute of Statistics and Decision Sciences Duke University, Durham NC 27706, USA. SUMMARY In Bayesian density estimation and

More information

Bayesian Learning. HT2015: SC4 Statistical Data Mining and Machine Learning. Maximum Likelihood Principle. The Bayesian Learning Framework

Bayesian Learning. HT2015: SC4 Statistical Data Mining and Machine Learning. Maximum Likelihood Principle. The Bayesian Learning Framework HT5: SC4 Statistical Data Mining and Machine Learning Dino Sejdinovic Department of Statistics Oxford http://www.stats.ox.ac.uk/~sejdinov/sdmml.html Maximum Likelihood Principle A generative model for

More information

Bayesian Inference and MCMC

Bayesian 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 information

Bayesian inference. Fredrik Ronquist and Peter Beerli. October 3, 2007

Bayesian inference. Fredrik Ronquist and Peter Beerli. October 3, 2007 Bayesian inference Fredrik Ronquist and Peter Beerli October 3, 2007 1 Introduction The last few decades has seen a growing interest in Bayesian inference, an alternative approach to statistical inference.

More information

Bayesian Inference for Regression Parameters

Bayesian Inference for Regression Parameters Bayesian Inference for Regression Parameters 1 Bayesian inference for simple linear regression parameters follows the usual pattern for all Bayesian analyses: 1. Form a prior distribution over all unknown

More information

Probabilistic Machine Learning

Probabilistic Machine Learning Probabilistic Machine Learning Bayesian Nets, MCMC, and more Marek Petrik 4/18/2017 Based on: P. Murphy, K. (2012). Machine Learning: A Probabilistic Perspective. Chapter 10. Conditional Independence Independent

More information

Bayesian Statistics Comparing Two Proportions or Means

Bayesian Statistics Comparing Two Proportions or Means Bayesian Statistics Comparing Two Proportions or Means Michael Anderson, PhD Hélène Carabin, DVM, PhD Department of Biostatistics and Epidemiology The University of Oklahoma Health Sciences Center May

More information

Prerequisite: STATS 7 or STATS 8 or AP90 or (STATS 120A and STATS 120B and STATS 120C). AP90 with a minimum score of 3

Prerequisite: STATS 7 or STATS 8 or AP90 or (STATS 120A and STATS 120B and STATS 120C). AP90 with a minimum score of 3 University of California, Irvine 2017-2018 1 Statistics (STATS) Courses STATS 5. Seminar in Data Science. 1 Unit. An introduction to the field of Data Science; intended for entering freshman and transfers.

More information

Bayesian Statistical Methods. Jeff Gill. Department of Political Science, University of Florida

Bayesian Statistical Methods. Jeff Gill. Department of Political Science, University of Florida Bayesian Statistical Methods Jeff Gill Department of Political Science, University of Florida 234 Anderson Hall, PO Box 117325, Gainesville, FL 32611-7325 Voice: 352-392-0262x272, Fax: 352-392-8127, Email:

More information

Bayesian Graphical Models

Bayesian Graphical Models Graphical Models and Inference, Lecture 16, Michaelmas Term 2009 December 4, 2009 Parameter θ, data X = x, likelihood L(θ x) p(x θ). Express knowledge about θ through prior distribution π on θ. Inference

More information

Metropolis Hastings. Rebecca C. Steorts Bayesian Methods and Modern Statistics: STA 360/601. Module 9

Metropolis Hastings. Rebecca C. Steorts Bayesian Methods and Modern Statistics: STA 360/601. Module 9 Metropolis Hastings Rebecca C. Steorts Bayesian Methods and Modern Statistics: STA 360/601 Module 9 1 The Metropolis-Hastings algorithm is a general term for a family of Markov chain simulation methods

More information

STA 4273H: Sta-s-cal Machine Learning

STA 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 information

Data Analysis and Uncertainty Part 2: Estimation

Data Analysis and Uncertainty Part 2: Estimation Data Analysis and Uncertainty Part 2: Estimation Instructor: Sargur N. University at Buffalo The State University of New York srihari@cedar.buffalo.edu 1 Topics in Estimation 1. Estimation 2. Desirable

More information

Bayesian Inference: Probit and Linear Probability Models

Bayesian Inference: Probit and Linear Probability Models Utah State University DigitalCommons@USU All Graduate Plan B and other Reports Graduate Studies 5-1-2014 Bayesian Inference: Probit and Linear Probability Models Nate Rex Reasch Utah State University Follow

More information

Bayesian Inference in GLMs. Frequentists typically base inferences on MLEs, asymptotic confidence

Bayesian 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 information

Bayesian Inference for DSGE Models. Lawrence J. Christiano

Bayesian 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 information

Bayesian Regression Linear and Logistic Regression

Bayesian 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 information

36-463/663: Hierarchical Linear Models

36-463/663: Hierarchical Linear Models 36-463/663: Hierarchical Linear Models Taste of MCMC / Bayes for 3 or more levels Brian Junker 132E Baker Hall brian@stat.cmu.edu 1 Outline Practical Bayes Mastery Learning Example A brief taste of JAGS

More information

Default Priors and Effcient Posterior Computation in Bayesian

Default 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 information

Bayesian Inference for DSGE Models. Lawrence J. Christiano

Bayesian 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 information

CHAPTER 10. Bayesian methods. Geir Storvik

CHAPTER 10. Bayesian methods. Geir Storvik CHAPTER 10 Bayesian methods Geir Storvik 10.1 Introduction Statistical inference concerns about learning from data, either parameters (estimation) or some, typically future, variables (prediction). In

More information

Bayesian data analysis in practice: Three simple examples

Bayesian 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 information

A Bayesian Approach for Sample Size Determination in Method Comparison Studies

A Bayesian Approach for Sample Size Determination in Method Comparison Studies A Bayesian Approach for Sample Size Determination in Method Comparison Studies Kunshan Yin a, Pankaj K. Choudhary a,1, Diana Varghese b and Steven R. Goodman b a Department of Mathematical Sciences b Department

More information

Part III. A Decision-Theoretic Approach and Bayesian testing

Part III. A Decision-Theoretic Approach and Bayesian testing Part III A Decision-Theoretic Approach and Bayesian testing 1 Chapter 10 Bayesian Inference as a Decision Problem The decision-theoretic framework starts with the following situation. We would like to

More information

The Bayesian Choice. Christian P. Robert. From Decision-Theoretic Foundations to Computational Implementation. Second Edition.

The 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 information

Bios 6649: Clinical Trials - Statistical Design and Monitoring

Bios 6649: Clinical Trials - Statistical Design and Monitoring Bios 6649: Clinical Trials - Statistical Design and Monitoring Spring Semester 2015 John M. Kittelson Department of Biostatistics & Informatics Colorado School of Public Health University of Colorado Denver

More information

Module 7: Introduction to Gibbs Sampling. Rebecca C. Steorts

Module 7: Introduction to Gibbs Sampling. Rebecca C. Steorts Module 7: Introduction to Gibbs Sampling Rebecca C. Steorts Agenda Gibbs sampling Exponential example Normal example Pareto example Gibbs sampler Suppose p(x, y) is a p.d.f. or p.m.f. that is difficult

More information

Advanced Statistical Modelling

Advanced Statistical Modelling Markov chain Monte Carlo (MCMC) Methods and Their Applications in Bayesian Statistics School of Technology and Business Studies/Statistics Dalarna University Borlänge, Sweden. Feb. 05, 2014. Outlines 1

More information

Linear Models A linear model is defined by the expression

Linear Models A linear model is defined by the expression Linear Models A linear model is defined by the expression x = F β + ɛ. where x = (x 1, x 2,..., x n ) is vector of size n usually known as the response vector. β = (β 1, β 2,..., β p ) is the transpose

More information

ST 740: Model Selection

ST 740: Model Selection ST 740: Model Selection Alyson Wilson Department of Statistics North Carolina State University November 25, 2013 A. Wilson (NCSU Statistics) Model Selection November 25, 2013 1 / 29 Formal Bayesian Model

More information

Metropolis-Hastings Algorithm

Metropolis-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 information

Comparison of Three Calculation Methods for a Bayesian Inference of Two Poisson Parameters

Comparison of Three Calculation Methods for a Bayesian Inference of Two Poisson Parameters Journal of Modern Applied Statistical Methods Volume 13 Issue 1 Article 26 5-1-2014 Comparison of Three Calculation Methods for a Bayesian Inference of Two Poisson Parameters Yohei Kawasaki Tokyo University

More information

Dynamic System Identification using HDMR-Bayesian Technique

Dynamic System Identification using HDMR-Bayesian Technique Dynamic System Identification using HDMR-Bayesian Technique *Shereena O A 1) and Dr. B N Rao 2) 1), 2) Department of Civil Engineering, IIT Madras, Chennai 600036, Tamil Nadu, India 1) ce14d020@smail.iitm.ac.in

More information

The STS Surgeon Composite Technical Appendix

The 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 information

Penalized Loss functions for Bayesian Model Choice

Penalized Loss functions for Bayesian Model Choice Penalized Loss functions for Bayesian Model Choice Martyn International Agency for Research on Cancer Lyon, France 13 November 2009 The pure approach For a Bayesian purist, all uncertainty is represented

More information

Part 8: GLMs and Hierarchical LMs and GLMs

Part 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 information

Bayesian Inference for DSGE Models. Lawrence J. Christiano

Bayesian 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 information

Introduction to Bayesian Statistics 1

Introduction to Bayesian Statistics 1 Introduction to Bayesian Statistics 1 STA 442/2101 Fall 2018 1 This slide show is an open-source document. See last slide for copyright information. 1 / 42 Thomas Bayes (1701-1761) Image from the Wikipedia

More information

Sequential Monte Carlo and Particle Filtering. Frank Wood Gatsby, November 2007

Sequential Monte Carlo and Particle Filtering. Frank Wood Gatsby, November 2007 Sequential Monte Carlo and Particle Filtering Frank Wood Gatsby, November 2007 Importance Sampling Recall: Let s say that we want to compute some expectation (integral) E p [f] = p(x)f(x)dx and we remember

More information

David B. Dahl. Department of Statistics, and Department of Biostatistics & Medical Informatics University of Wisconsin Madison

David B. Dahl. Department of Statistics, and Department of Biostatistics & Medical Informatics University of Wisconsin Madison AN IMPROVED MERGE-SPLIT SAMPLER FOR CONJUGATE DIRICHLET PROCESS MIXTURE MODELS David B. Dahl dbdahl@stat.wisc.edu Department of Statistics, and Department of Biostatistics & Medical Informatics University

More information

CSC 2541: Bayesian Methods for Machine Learning

CSC 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 information

Markov Chain Monte Carlo in Practice

Markov 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 information

Non-Parametric Bayes

Non-Parametric Bayes Non-Parametric Bayes Mark Schmidt UBC Machine Learning Reading Group January 2016 Current Hot Topics in Machine Learning Bayesian learning includes: Gaussian processes. Approximate inference. Bayesian

More information

Bayesian RL Seminar. Chris Mansley September 9, 2008

Bayesian RL Seminar. Chris Mansley September 9, 2008 Bayesian RL Seminar Chris Mansley September 9, 2008 Bayes Basic Probability One of the basic principles of probability theory, the chain rule, will allow us to derive most of the background material in

More information

Introduction 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 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 information

DAG models and Markov Chain Monte Carlo methods a short overview

DAG models and Markov Chain Monte Carlo methods a short overview DAG models and Markov Chain Monte Carlo methods a short overview Søren Højsgaard Institute of Genetics and Biotechnology University of Aarhus August 18, 2008 Printed: August 18, 2008 File: DAGMC-Lecture.tex

More information

Hierarchical Linear Models

Hierarchical Linear Models Hierarchical Linear Models Statistics 220 Spring 2005 Copyright c 2005 by Mark E. Irwin The linear regression model Hierarchical Linear Models y N(Xβ, Σ y ) β σ 2 p(β σ 2 ) σ 2 p(σ 2 ) can be extended

More information