DOCSLIB.ORG

Marginal Likelihood Estimation Via Power Posteriors

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Marginal Likelihood Estimation Via Power Posteriors Marginal likelihood estimation via power posteriors N. Friel† University of Glasgow, UK A.N. Pettitt Queensland University of Technology, Australia Summary. Model choice plays an increasingly important role in Statistics. From a Bayesian perspective a crucial goal is to compute the marginal likelihood of the data for a given model. This however is typically a difficult task since it amounts to integrating over all model parameters. The aim of this paper is to illustrate how this may be achieved using ideas from thermodynamic integration or path sampling. We show how the marginal likelihood can be computed via MCMC methods on modified posterior distributions for each model. This then allows Bayes factors or posterior model probabilitiesto be calculated. We show that this approach requires very little tuning, and is straightforward to implement. The new method is illustrated in a variety of challenging statistical settings. Keywords: BAYES FACTOR; REGRESSION; LONGITUDINAL DATA; MULTINOMIAL; HIDDEN MARKOV MODEL; MODEL CHOICE 1. Introduction Suppose data y is assumed to have been generated by one of M models indexed by the set {1, 2,...,M}. An important goal of Bayesian model selection is to calculate p(k|y) – the posterior model probability for model k. Here the aim may be to obtain a single most probable model, or indeed a subset of likely models, a posteriori. Alternatively, posterior model probabilities may be synthesised from all competing models to calculate some quantity of interest common to all models, using model averaging (Hoeting et al 2001). But before we can make inference for p(k|y) we first need to fully specify the posterior distribution of all unknown parameters in the joint model and parameter space. We denote by θk, the parameters specific to model k, where θ denotes collection of all model parameters. Specifying prior distributions for within model parameters p(θk|k), and for model indi- cators p(k), together with a likelihood L(y|θk, k) within model k allows Bayesian inference to proceed by examining the posterior distribution p(θk, k|y) ∝ L(y|θk, k)p(θk|k)p(k). (1) Across model strategies proceed by sampling from this joint posterior distribution of model indicators and parameters. The reversible jump MCMC algorithm (Green 1995) is a popular approach for this situation. Other across model search strategies include (Godsill 2001) and (Carlin and Chib 1995). By contrast, within model searches examine the posterior distribution within model k separately for each k. Here the within model posterior appears as p(θk|y, k) ∝ L(y|θk, k)p(θk|k), where the constant of proportionality, often termed the marginal likelihood or integrated likelihood for model k is written as p(y|k)= L(y|θk, k)p(θk|k)dθk. (2) Zθk This is in general a difficult integral to compute, possibly involving high dimensional within model parameters θk. However if we were able to do so, then we would be readily able to make statements about posterior model probabilities, using Bayes theorem, p(y|k)p(k) p(k|y)= M . 1 p(y|k)p(k) †Address for correspondence: Department of Statistics,P University of Glasgow, Glasgow, G12 8QQ, UK. E-mail: [email protected] 2 Friel, Pettitt The marginal likelihoods can be used to compare two models by computing Bayes factors, p(y|k = 1) B = , 12 p(y|k = 2) without the need to specify prior model probabilities. Note that the Bayes factor B12 gives the evidence provided by the data in favour of model 1 compared to model 2. It can also be seen that p(k =1|y) p(k = 2) B = . 12 p(k =2|y) p(k = 1) In other words, the Bayes factor is the ratio of the posterior odds of model 1 to its prior odds. Note that an improper prior distribution p(θk|k) leads necessarily to an improper marginal like- lihood, which in turns implies that the Bayes factor is not well-defined. To circumvent the difficulty of using improper priors for model comparison, (O’Hagan 1995) introduced an approximate method termed the fractional Bayes factors. Here an approximate (proper) marginal likelihood is defined by the ratio L(y|θ , k)p(θ |k)dθ θk k k k , R {L(y|θ , k)}ap(θ |k)dθ θk k k k R since any impropriety in the prior for θk cancels above and below. Other approaches to this problem include the intrinsic Bayes factor (Berger and Perricchi 1996). In this paper we concentrate on the case where prior model distributions are proper. Various methods have been proposed in the literature to estimate the marginal likelihood (2). For example (Chib 1995) estimates the marginal likelihood p(y|k) using output from a Gibbs sampler for (θk|y, k). It relies however on a block updating approach for θk. Clearly this is not always possible to do. This work was then extended in (Chib and Jeliazkov 2001), where output from a Metropolis-Hasting algorithm for the posterior (θk|y, k) can be used to estimate the marginal likelihood. Annealed importance sampling (Neal 2001) estimates the marginal likelihood using ideas from importance sampling. Here an independent sample from the posterior is generated by defining a sequence of distributions indexed by a temperature parameter t from the prior through to the posterior. Importantly however, the collection of importance weights can be used to estimate the marginal likelihood. In this paper we propose a new method to compute the marginal likelihood based on samples from a distribution proportional to the likelihood raised to a power t times the prior, which we term the power posterior. This method was inspired by ideas from path sampling or thermodynamic integration (Gelman and Meng 1998). We find that the marginal likelihood, p(y|k) can be expressed as an integral with respect to t from 0 to 1 of the expected deviance for model k, where the expectation is taken with respect to the power distribution, at power t. We argue that this method requires very little tuning or book keeping. It is easy to implement, requiring minor modification to computer code which samples from the posterior distribution. In this paper, in Section 2 we carry out a review of different approaches to across and within model searches. Section 3 introduces the new method for estimating the marginal likelihood, based on sampling from the so-called power posterior distribution. Here we also outline how this method could be implemented in practice, while also giving some guidance as to the sensitivity of the estimate of the marginal likelihood (or Bayes factor) to the diffusivity of the prior model parameters. Three illustrations of how the new method performs in practice are given in Section 4. In particular complex modelling situations are illustrated which preclude other methods of marginal likelihood estimation which rely on block updating model parameters. We conclude this paper with some final remarks in Section 5. 2. Review of methods to compute Bayes factors There are numerous methods and techniques available to estimate (2). Generally speaking two ap- proaches are possible – an across model search or a within model search. The former approach in an MCMC setting, involves generating a single Markov chain which traverses the joint model and parameter space (2). A popular choice is the reversible jump sampler (Green 1995). Godsill (2001) retains aspects of the reversible jump sampler, but considers the case where parameters are shared Marginal likelihood estimation via power posteriors 3 between different model, as occurs for example in nested models. Other choices include (Stephens 2000), (Carlin and Chib 1995) and (Dellaportas, Forster and Ntzoufras 2001). Within model search essentially aims to estimate the marginal likelihood (2) for each model k separately, and then if desired uses this information to form Bayes factors, see (Chib 1995), (Chib and Jeliazkov 2001). Neal (2001) combines aspects of simulated annealing and importance sampling to provide a method of gathering an independent sample from a posterior distribution of interest, but importantly also to estimate the marginal likelihood. Bridge sampling (Meng and Wong 1996) offers the possibility of estimating the Bayes factor linking the two posterior distributions by a bridge function. Bartolucci and Mira (2003) use this approach, although it is based on an across model reversible jump sampler. Within model approaches are disadvantageous when the cardinality of the model space is large. However, as noted by Green (2003), the ideal situation for a within model approach is one where the models are all reasonably heterogeneous. In effect, this is the case where it is difficult to choose proposal distributions when jumping between models - and indeed the situation where parameters across models of the same dimension have different interpretations. This short review is not intended to be exhaustive. A more complete picture can be found in the substantial reviews (Sisson 2005) and (Han and Carlin 2001). 2.1. Reversible jump MCMC Reversible jump MCMC (Green 1995) offers the potential to carry out inference for all unknown parameters (1) in the joint model and parameter space in a single logical framework. A crucial innovation in the seminal paper by Green (1995) was to illustrate that detailed balance could be achieved for general state spaces. In particular this extends the Metropolis-Hastings algorithm to variable dimension state spaces of the type (θk, k). To implement the algorithm, proposing to move from (θk, k) to (θl,l) proceeds by generating a random vector u from a distribution g and setting ∗ (θl,l)= fkl((θk, k), u). Similarly to move from (θl,l) to (θk, k) requires random numbers u follow- ∗ ∗ ing some distribution g , and setting (θk, k) = flk((θl,l), u ), for some deterministic function flk.
Load more

Recommended publications
  • The Composite Marginal Likelihood (CML) Inference Approach with Applications to Discrete and Mixed Dependent Variable Models
    Technical Report 101 The Composite Marginal Likelihood (CML) Inference Approach with Applications to Discrete and Mixed Dependent Variable Models Chandra R. Bhat Center for Transportation Research September 2014 Data-Supported Transportation Operations & Planning Center (D-STOP) A Tier 1 USDOT University Transportation Center at The University of Texas at Austin D-STOP is a collaborative initiative by researchers at the Center for Transportation Research and the Wireless Networking and Communications Group at The University of Texas at Austin. DISCLAIMER The contents of this report reflect the views of the authors, who are responsible for the facts and the accuracy of the information presented herein. This document is disseminated under the sponsorship of the U.S. Department of Transportation’s University Transportation Centers Program, in the interest of information exchange. The U.S. Government assumes no liability for the contents or use thereof. Technical Report Documentation Page 1. Report No. 2. Government Accession No. 3. Recipient's Catalog No. D-STOP/2016/101 4. Title and Subtitle 5. Report Date The Composite Marginal Likelihood (CML) Inference Approach September 2014 with Applications to Discrete and Mixed Dependent Variable 6. Performing Organization Code Models 7. Author(s) 8. Performing Organization Report No. Chandra R. Bhat Report 101 9. Performing Organization Name and Address 10. Work Unit No. (TRAIS) Data-Supported Transportation Operations & Planning Center (D- STOP) 11. Contract or Grant No. The University of Texas at Austin DTRT13-G-UTC58 1616 Guadalupe Street, Suite 4.202 Austin, Texas 78701 12. Sponsoring Agency Name and Address 13. Type of Report and Period Covered Data-Supported Transportation Operations & Planning Center (D- STOP) The University of Texas at Austin 14.
    [Show full text]
  • A Widely Applicable Bayesian Information Criterion
    JournalofMachineLearningResearch14(2013)867-897 Submitted 8/12; Revised 2/13; Published 3/13 A Widely Applicable Bayesian Information Criterion Sumio Watanabe [email protected] Department of Computational Intelligence and Systems Science Tokyo Institute of Technology Mailbox G5-19, 4259 Nagatsuta, Midori-ku Yokohama, Japan 226-8502 Editor: Manfred Opper Abstract A statistical model or a learning machine is called regular if the map taking a parameter to a prob- ability distribution is one-to-one and if its Fisher information matrix is always positive definite. If otherwise, it is called singular. In regular statistical models, the Bayes free energy, which is defined by the minus logarithm of Bayes marginal likelihood, can be asymptotically approximated by the Schwarz Bayes information criterion (BIC), whereas in singular models such approximation does not hold. Recently, it was proved that the Bayes free energy of a singular model is asymptotically given by a generalized formula using a birational invariant, the real log canonical threshold (RLCT), instead of half the number of parameters in BIC. Theoretical values of RLCTs in several statistical models are now being discovered based on algebraic geometrical methodology. However, it has been difficult to estimate the Bayes free energy using only training samples, because an RLCT depends on an unknown true distribution. In the present paper, we define a widely applicable Bayesian information criterion (WBIC) by the average log likelihood function over the posterior distribution with the inverse temperature 1/logn, where n is the number of training samples. We mathematically prove that WBIC has the same asymptotic expansion as the Bayes free energy, even if a statistical model is singular for or unrealizable by a statistical model.
    [Show full text]
  • Categorical Distributions in Natural Language Processing Version 0.1
    Categorical Distributions in Natural Language Processing Version 0.1 MURAWAKI Yugo 12 May 2016 MURAWAKI Yugo Categorical Distributions in NLP 1 / 34 Categorical distribution Suppose random variable x takes one of K values. x is generated according to categorical distribution Cat(θ), where θ = (0:1; 0:6; 0:3): RYG In many task settings, we do not know the true θ and need to infer it from observed data x = (x1; ··· ; xN). Once we infer θ, we often want to predict new variable x0. NOTE: In Bayesian settings, θ is usually integrated out and x0 is predicted directly from x. MURAWAKI Yugo Categorical Distributions in NLP 2 / 34 Categorical distributions are a building block of natural language models N-gram language model (predicting the next word) POS tagging based on a Hidden Markov Model (HMM) Probabilistic context-free grammar (PCFG) Topic model (Latent Dirichlet Allocation (LDA)) MURAWAKI Yugo Categorical Distributions in NLP 3 / 34 Example: HMM-based POS tagging BOS DT NN VBZ VBN EOS the sun has risen Let K be the number of POS tags and V be the vocabulary size (ignore BOS and EOS for simplicity). The transition probabilities can be computed using K categorical θTRANS; θTRANS; ··· distributions ( DT NN ), with the dimension K. θTRANS = : ; : ; : ; ··· DT (0 21 0 27 0 09 ) NN NNS ADJ Similarly, the emission probabilities can be computed using K θEMIT; θEMIT; ··· categorical distributions ( DT NN ), with the dimension V. θEMIT = : ; : ; : ; ··· NN (0 012 0 002 0 005 ) sun rose cat MURAWAKI Yugo Categorical Distributions in NLP 4 / 34 Outline Categorical and multinomial distributions Conjugacy and posterior predictive distribution LDA (Latent Dirichlet Applocation) as an application Gibbs sampling for inference MURAWAKI Yugo Categorical Distributions in NLP 5 / 34 Categorical distribution: 1 observation Suppose θ is known.
    [Show full text]
  • Binomial and Multinomial Distributions
    Binomial and multinomial distributions Kevin P. Murphy Last updated October 24, 2006 * Denotes more advanced sections 1 Introduction In this chapter, we study probability distributions that are suitable for modelling discrete data, like letters and words. This will be useful later when we consider such tasks as classifying and clustering documents, recognizing and segmenting languages and DNA sequences, data compression, etc. Formally, we will be concerned with density models for X ∈{1,...,K}, where K is the number of possible values for X; we can think of this as the number of symbols/ letters in our alphabet/ language. We assume that K is known, and that the values of X are unordered: this is called categorical data, as opposed to ordinal data, in which the discrete states can be ranked (e.g., low, medium and high). If K = 2 (e.g., X represents heads or tails), we will use a binomial distribution. If K > 2, we will use a multinomial distribution. 2 Bernoullis and Binomials Let X ∈{0, 1} be a binary random variable (e.g., a coin toss). Suppose p(X =1)= θ. Then − p(X|θ) = Be(X|θ)= θX (1 − θ)1 X (1) is called a Bernoulli distribution. It is easy to show that E[X] = p(X =1)= θ (2) Var [X] = θ(1 − θ) (3) The likelihood for a sequence D = (x1,...,xN ) of coin tosses is N − p(D|θ)= θxn (1 − θ)1 xn = θN1 (1 − θ)N0 (4) n=1 Y N N where N1 = n=1 xn is the number of heads (X = 1) and N0 = n=1(1 − xn)= N − N1 is the number of tails (X = 0).
    [Show full text]
  • Bayesian Monte Carlo
    Bayesian Monte Carlo Carl Edward Rasmussen and Zoubin Ghahramani Gatsby Computational Neuroscience Unit University College London 17 Queen Square, London WC1N 3AR, England edward,[email protected] http://www.gatsby.ucl.ac.uk Abstract We investigate Bayesian alternatives to classical Monte Carlo methods for evaluating integrals. Bayesian Monte Carlo (BMC) allows the in- corporation of prior knowledge, such as smoothness of the integrand, into the estimation. In a simple problem we show that this outperforms any classical importance sampling method. We also attempt more chal- lenging multidimensional integrals involved in computing marginal like- lihoods of statistical models (a.k.a. partition functions and model evi- dences). We find that Bayesian Monte Carlo outperformed Annealed Importance Sampling, although for very high dimensional problems or problems with massive multimodality BMC may be less adequate. One advantage of the Bayesian approach to Monte Carlo is that samples can be drawn from any distribution. This allows for the possibility of active design of sample points so as to maximise information gain. 1 Introduction Inference in most interesting machine learning algorithms is not computationally tractable, and is solved using approximations. This is particularly true for Bayesian models which require evaluation of complex multidimensional integrals. Both analytical approximations, such as the Laplace approximation and variational methods, and Monte Carlo methods have recently been used widely for Bayesian machine learning problems. It is interesting to note that Monte Carlo itself is a purely frequentist procedure [O’Hagan, 1987; MacKay, 1999]. This leads to several inconsistencies which we review below, outlined in a paper by O’Hagan [1987] with the title “Monte Carlo is Fundamentally Unsound”.
    [Show full text]
  • Marginal Likelihood
    STA 4273H: Statistical Machine Learning Russ Salakhutdinov Department of Statistics! [email protected]! http://www.utstat.utoronto.ca/~rsalakhu/ Sidney Smith Hall, Room 6002 Lecture 2 Last Class •" In our last class, we looked at: -" Statistical Decision Theory -" Linear Regression Models -" Linear Basis Function Models -" Regularized Linear Regression Models -" Bias-Variance Decomposition •" We will now look at the Bayesian framework and Bayesian Linear Regression Models. Bayesian Approach •" We formulate our knowledge about the world probabilistically: -" We define the model that expresses our knowledge qualitatively (e.g. independence assumptions, forms of distributions). -" Our model will have some unknown parameters. -" We capture our assumptions, or prior beliefs, about unknown parameters (e.g. range of plausible values) by specifying the prior distribution over those parameters before seeing the data. •" We observe the data. •" We compute the posterior probability distribution for the parameters, given observed data. •" We use this posterior distribution to: -" Make predictions by averaging over the posterior distribution -" Examine/Account for uncertainly in the parameter values. -" Make decisions by minimizing expected posterior loss. (See Radford Neal’s NIPS tutorial on ``Bayesian Methods for Machine Learning'’) Posterior Distribution •" The posterior distribution for the model parameters can be found by combining the prior with the likelihood for the parameters given the data. •" This is accomplished using Bayes’
    [Show full text]
  • Bayesian Inference
    Bayesian Inference Thomas Nichols With thanks Lee Harrison Bayesian segmentation Spatial priors Posterior probability Dynamic Causal and normalisation on activation extent maps (PPMs) Modelling Attention to Motion Paradigm Results SPC V3A V5+ Attention – No attention Büchel & Friston 1997, Cereb. Cortex Büchel et al. 1998, Brain - fixation only - observe static dots + photic V1 - observe moving dots + motion V5 - task on moving dots + attention V5 + parietal cortex Attention to Motion Paradigm Dynamic Causal Models Model 1 (forward): Model 2 (backward): attentional modulation attentional modulation of V1→V5: forward of SPC→V5: backward Photic SPC Attention Photic SPC V1 V1 - fixation only V5 - observe static dots V5 - observe moving dots Motion Motion - task on moving dots Attention Bayesian model selection: Which model is optimal? Responses to Uncertainty Long term memory Short term memory Responses to Uncertainty Paradigm Stimuli sequence of randomly sampled discrete events Model simple computational model of an observers response to uncertainty based on the number of past events (extent of memory) 1 2 3 4 Question which regions are best explained by short / long term memory model? … 1 2 40 trials ? ? Overview • Introductory remarks • Some probability densities/distributions • Probabilistic (generative) models • Bayesian inference • A simple example – Bayesian linear regression • SPM applications – Segmentation – Dynamic causal modeling – Spatial models of fMRI time series Probability distributions and densities k=2 Probability distributions
    [Show full text]
  • On the Derivation of the Bayesian Information Criterion
    On the derivation of the Bayesian Information Criterion H. S. Bhat∗† N. Kumar∗ November 8, 2010 Abstract We present a careful derivation of the Bayesian Inference Criterion (BIC) for model selection. The BIC is viewed here as an approximation to the Bayes Factor. One of the main ingredients in the approximation, the use of Laplace’s method for approximating integrals, is explained well in the literature. Our derivation sheds light on this and other steps in the derivation, such as the use of a flat prior and the invocation of the weak law of large numbers, that are not often discussed in detail. 1 Notation Let us define the notation that we will use: y : observed data y1,...,yn Mi : candidate model P (y|Mi) : marginal likelihood of the model Mi given the data θi : vector of parameters in the model Mi gi(θi) : the prior density of the parameters θi f(y|θi) : the density of the data given the parameters θi L(θi|y) : the likelihood of y given the model Mi θˆi : the MLE of θi that maximizes L(θi|y) 2 Bayes Factor The Bayesian approach to model selection [1] is to maximize the posterior probability of n a model (Mi) given the data {yj}j=1. Applying Bayes theorem to calculate the posterior ∗School of Natural Sciences, University of California, Merced, 5200 N. Lake Rd., Merced, CA 95343 †Corresponding author, email: [email protected] 1 probability of a model given the data, we get P (y1,...,yn|Mi)P (Mi) P (Mi|y1,...,yn)= , (1) P (y1,...,yn) where P (y1,...,yn|Mi) is called the marginal likelihood of the model Mi.
    [Show full text]
  • Marginal Likelihood from the Gibbs Output Siddhartha Chib Journal Of
    Marginal Likelihood from the Gibbs Output Siddhartha Chib Journal of the American Statistical Association, Vol. 90, No. 432. (Dec., 1995), pp. 1313-1321. Stable URL: http://links.jstor.org/sici?sici=0162-1459%28199512%2990%3A432%3C1313%3AMLFTGO%3E2.0.CO%3B2-2 Journal of the American Statistical Association is currently published by American Statistical Association. Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/journals/astata.html. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. The JSTOR Archive is a trusted digital repository providing for long-term preservation and access to leading academic journals and scholarly literature from around the world. The Archive is supported by libraries, scholarly societies, publishers, and foundations. It is an initiative of JSTOR, a not-for-profit organization with a mission to help the scholarly community take advantage of advances in technology. For more information regarding JSTOR, please contact [email protected]. http://www.jstor.org Wed Feb 6 14:39:37 2008 Marginal Likelihood From the Gibbs Output Siddhartha CHIB In the context of Bayes estimation via Gibbs sampling, with or without data augmentation, a simple approach is developed for computing the marginal density of the sample data (marginal likelihood) given parameter draws from the posterior distribution.
    [Show full text]
  • Bayesian Analysis of Graphical Models of Marginal Independence for Three Way Contingency Tables
    A Service of Leibniz-Informationszentrum econstor Wirtschaft Leibniz Information Centre Make Your Publications Visible. zbw for Economics Tarantola, Claudia; Ntzoufras, Ioannis Working Paper Bayesian Analysis of Graphical Models of Marginal Independence for Three Way Contingency Tables Quaderni di Dipartimento, No. 172 Provided in Cooperation with: University of Pavia, Department of Economics and Quantitative Methods (EPMQ) Suggested Citation: Tarantola, Claudia; Ntzoufras, Ioannis (2012) : Bayesian Analysis of Graphical Models of Marginal Independence for Three Way Contingency Tables, Quaderni di Dipartimento, No. 172, Università degli Studi di Pavia, Dipartimento di Economia Politica e Metodi Quantitativi (EPMQ), Pavia This Version is available at: http://hdl.handle.net/10419/95304 Standard-Nutzungsbedingungen: Terms of use: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Documents in EconStor may be saved and copied for your Zwecken und zum Privatgebrauch gespeichert und kopiert werden. personal and scholarly purposes. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle You are not to copy documents for public or commercial Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich purposes, to exhibit the documents publicly, to make them machen, vertreiben oder anderweitig nutzen. publicly available on the internet, or to distribute or otherwise use the documents in public. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten,
    [Show full text]
  • Objective Priors in the Empirical Bayes Framework Arxiv:1612.00064V5 [Stat.ME] 11 May 2020
    Objective Priors in the Empirical Bayes Framework Ilja Klebanov1 Alexander Sikorski1 Christof Sch¨utte1,2 Susanna R¨oblitz3 May 13, 2020 Abstract: When dealing with Bayesian inference the choice of the prior often re- mains a debatable question. Empirical Bayes methods offer a data-driven solution to this problem by estimating the prior itself from an ensemble of data. In the non- parametric case, the maximum likelihood estimate is known to overfit the data, an issue that is commonly tackled by regularization. However, the majority of regu- larizations are ad hoc choices which lack invariance under reparametrization of the model and result in inconsistent estimates for equivalent models. We introduce a non-parametric, transformation invariant estimator for the prior distribution. Being defined in terms of the missing information similar to the reference prior, it can be seen as an extension of the latter to the data-driven setting. This implies a natural interpretation as a trade-off between choosing the least informative prior and incor- porating the information provided by the data, a symbiosis between the objective and empirical Bayes methodologies. Keywords: Parameter estimation, Bayesian inference, Bayesian hierarchical model- ing, invariance, hyperprior, MPLE, reference prior, Jeffreys prior, missing informa- tion, expected information 2010 Mathematics Subject Classification: 62C12, 62G08 1 Introduction Inferring a parameter θ 2 Θ from a measurement x 2 X using Bayes' rule requires prior knowledge about θ, which is not given in many applications. This has led to a lot of controversy arXiv:1612.00064v5 [stat.ME] 11 May 2020 in the statistical community and to harsh criticism concerning the objectivity of the Bayesian approach.
    [Show full text]
  • Estimating the Marginal Likelihood with Integrated Nested Laplace Approximation (INLA)
    Estimating the marginal likelihood with Integrated nested Laplace approximation (INLA) Aliaksandr Hubin ∗ Department of Mathematics, University of Oslo and Geir Storvik Department of Mathematics, University of Oslo November 7, 2016 Abstract The marginal likelihood is a well established model selection criterion in Bayesian statistics. It also allows to efficiently calculate the marginal posterior model proba- bilities that can be used for Bayesian model averaging of quantities of interest. For many complex models, including latent modeling approaches, marginal likelihoods are however difficult to compute. One recent promising approach for approximating the marginal likelihood is Integrated Nested Laplace Approximation (INLA), design for models with latent Gaussian structures. In this study we compare the approxima- tions obtained with INLA to some alternative approaches on a number of examples of different complexity. In particular we address a simple linear latent model, a Bayesian linear regression model, logistic Bayesian regression models with probit and logit links, and a Poisson longitudinal generalized linear mixed model. Keywords: Integrated nested Laplace approximation; Marginal likelihood; Model Evidence; Bayes Factor; Markov chain Monte Carlo; Numerical Integration; Linear models; General- arXiv:1611.01450v1 [stat.CO] 4 Nov 2016 ized linear models; Generalized linear mixed models; Bayesian model selection; Bayesian model averaging. ∗The authors gratefully acknowledge the CELS project at the University of Oslo, http://www.mn.uio. no/math/english/research/groups/cels/index.html, for giving us the opportunity, inspiration and motivation to write this article. 1 1 Introduction Marginal likelihoods have been commonly accepted to be an extremely important quan- tity within Bayesian statistics. For data y and model M, which includes some unknown parameters θ, the marginal likelihood is given by Z p(yjM) = p(yjM; θ)p(θjM)dθ (1) Ωθ where p(θjM) is the prior for θ under model M while p(yjM; θ) is the likelihood function conditional on θ.
    [Show full text]