Dissertations / Theses on the topic 'Jeffreys prior'

This paper discusses characteristics of standard conjugate priors and their induced posteriors in Bayesian inference for von Mises-Fisher distributions, using either the canonical natural exponential family or the more commonly employed polar coordinate parameterizations. We analyze when standard conjugate priors as well as posteriors are proper, and investigate the Jeffreys prior for the von Mises-Fisher family. Finally, we characterize the proper distributions in the standard conjugate family of the (matrixvalued) von Mises-Fisher distributions on Stiefel manifolds.

Le but de cette thèse est d’étudier l’approximation d’a priori impropres par des suites d’a priori propres. Nous définissons un mode de convergence sur les mesures de Radon strictement positives pour lequel une suite de mesures de probabilité peut admettre une mesure impropre pour limite. Ce mode de convergence, que nous appelons convergence q-vague, est indépendant du modèle statistique. Il permet de comprendre l’origine du paradoxe de Jeffreys-Lindley. Ensuite, nous nous intéressons à l’estimation de la taille d’une population. Nous considérons le modèle du removal sampling. Nous établissons des conditions nécessaires et suffisantes sur un certain type d’a priori pour obtenir des estimateurs a posteriori bien définis. Enfin, nous montrons à l’aide de la convergence q-vague, que l’utilisation d’a priori vagues n’est pas adaptée car les estimateurs obtenus montrent une grande dépendance aux hyperparamètres
The purpose of this thesis is to study the approximation of improper priors by proper priors. We define a convergence mode on the positive Radon measures for which a sequence of probability measures could converge to an improper limiting measure. This convergence mode, called q-vague convergence, is independant from the statistical model. It explains the origin of the Jeffreys-Lindley paradox. Then, we focus on the estimation of the size of a population. We consider the removal sampling model. We give necessary and sufficient conditions on the hyperparameters in order to have proper posterior distributions and well define estimate of abundance. In the light of the q-vague convergence, we show that the use of vague priors is not appropriate in removal sampling since the estimates obtained depend crucially on hyperparameters

Orientador: Caio Lucidius Naberezny Azevedo
Dissertação (mestrado) - Universidade Estadual de Campinas, Instituto de Matemática Estatística e Computação Científica
Made available in DSpace on 2018-08-23T07:11:52Z (GMT). No. of bitstreams: 1 Nogarotto_DaniloCovaes_M.pdf: 12817108 bytes, checksum: 0e5e0de542d707f4023f5ef62dc40a82 (MD5) Previous issue date: 2013
Resumo: No presente trabalho desenvolvemos ferramentas de inferência bayesiana para modelos de regressão beta e beta inflacionados, em relação à estimação paramétrica e diagnóstico. Trabalhamos com modelos de regressão beta não inflacionados, inflacionados em zero ou um e inflacionados em zero e um. Devido à impossibilidade de obtenção analítica das posteriores de interesse, tais ferramentas foram desenvolvidas através de algoritmos MCMC. Para os parâmetros da estrutura de regressão e para o parâmetro de precisão exploramos a utilização de prioris comumente empregadas em modelos de regressão, bem como prioris de Jeffreys e de Jeffreys sob independência. Para os parâmetros das componentes discretas, consideramos prioris conjugadas. Realizamos diversos estudos de simulação considerando algumas situações de interesse prático com o intuito de comparar as estimativas bayesianas com as frequentistas e também de estudar a sensibilidade dos modelos _a escolha de prioris. Um conjunto de dados da área psicométrica foi analisado para ilustrar o potencial do ferramental desenvolvido. Os resultados indicaram que há ganho ao se considerar modelos que contemplam as observações inflacionadas ao invés de transformá-las a fim de utilizar modelos não inflacionados
Abstract: In the present work we developed Bayesian tools, concerning parameter estimation and diagnostics, for noninflated, zero inflated, one inflated and zero-one inflated beta regression models. Due to the impossibility of obtaining the posterior distributions of interest, analytically, all these tools were developed through MCMC algorithms. For the regression and precision parameters we exploited the using of prior distributions commonly considered in regression models as well as Jeffreys and independence Jeffreys priors. For the parameters related to the discrete components, we considered conjugate prior distributions. We performed simulation studies, considering some situations of practical interest, in order to compare the Bayesian and frequentist estimates as well as to evaluate the sensitivity of the models to the prior choice. A psychometric real data set was analyzed to illustrate the performance of the developed tools. The results indicated that there is an overall improvement in using models that consider the inflated observations compared to transforming these observations in order to use noninflated models
Mestrado
Estatistica
Mestre em Estatística

The Impact of Vintage on the Persistence of Gross Domestic Product Shocks. The first chapter of the thesis aims to demonstrate that the data revision process affects the persistence of gross domestic product shocks. The analysis is based on two alternative models, the Fractional Unit Root and the Linear Trend, and it benefits from new semiparametric procedures. The analysis of results seems to suggest that changes in the definition of the output significantly affects the performance of the models which are typically used to study the GDP series. - Seasonality in HighFrequency Data. The second chapter of the thesis aims to study the intradaily seasonal pattern of the Dow Jones volatility. An unobserved component analysis, following the socalled ModelBased approach, is examined in order to separate the seasonal pattern from the remaining long and short term components. The major novelty of the work is to explore the seasonal behavior of the volatility as a stochastic component which evolves over time according to a specific ARMA model. In more detail, high frequency data are used to recover 30minute realized volatility measures. Particular attention has been devoted to checking whether estimates were robust to market microstructure, jumps and outliers. The analysis of results emphasizes that the volatility of the Dow Jones is characterized by a stochastic seasonal component which recalls the “Ushape” pattern. - Bayesian Unobserved Components in Time Series. The third chapter of the thesis aims to present a full Bayesian framework to identify, extract and forecast unobserved components in time series. The major novelty of the approach is the definition of a probabilistic framework to analyze the identification conditions. More precisely, informative prior distributions are assigned to the spectral densities of the unobserved components. This entails a interesting feature: the possibility to analyze more than one decomposition at once by studying the posterior distributions of the unobserved spectra. Particular attention is given to an empirical application where the canonical decomposition of sunspot data is compared with some alternative decompositions. The posterior distributions of the unobserved components are recovered by exploiting some recent developments in the WienerKolmogorov and circular process literature. An empirical application shows how to capture the seasonal component in the volatility of financial high frequency data. The posterior forecasting distributions are finally recovered by exploiting a relationship between spectral densities and linear processes. An empirical application shows how to forecast seasonal adjusted financial time series. Finally, a generalization of the BernsteinDirichlet prior distribution is proposed in order to implement a frequencypass spectral density estimator. - Objective Priors for AR(p) models. The fourth chapter of the thesis aims to derive objective prior distributions for general autoregressive models. The core of the paper is based on the study of the Jeffreys' principle and on a particular adaptation of the “reference” algorithm for dependent data. The analysis of stationarity turns out to be rather complicated for general autoregressive processes. Therefore, two alternative parameterizations which respectively depend on the partial autocorrelation functions (PACFs) and the roots are considered. For the causal and stationary parameter subspace, the PACFs and the roots are always defined within the unit circle. This simplifies the derivation of the prior distributions. Two main results are obtained. The first is a general formula which depends on the PACFs and represents the Jeffreys' prior distribution for the causal subspace. The second result depends on the roots of the process and represents a particular reference prior distribution which is asymptotically independent from the initial conditions and covers the noncausal subspace. Ultimately, simulation results are presented for the autoregressive process of order two.

Récemment, la grande complexité des applications modernes, par exemple dans la génétique, l’informatique, la finance, les sciences du climat, etc. a conduit à la proposition des nouveaux modèles qui peuvent décrire la réalité. Dans ces cas,méthodes MCMC classiques ne parviennent pas à rapprocher la distribution a posteriori, parce qu’ils sont trop lents pour étudier le space complet du paramètre. Nouveaux algorithmes ont été proposés pour gérer ces situations, où la fonction de vraisemblance est indisponible. Nous allons étudier nombreuses caractéristiques des modèles complexes: comment éliminer les paramètres de nuisance de l’analyse et faire inférence sur les quantités d’intérêt,dans un cadre bayésienne et non bayésienne et comment construire une distribution a priori de référence
Recently, the great complexity of modern applications, for instance in genetics,computer science, finance, climatic science etc., has led to the proposal of newmodels which may realistically describe the reality. In these cases, classical MCMCmethods fail to approximate the posterior distribution, because they are too slow toinvestigate the full parameter space. New algorithms have been proposed to handlethese situations, where the likelihood function is unavailable. We will investigatemany features of complex models: how to eliminate the nuisance parameters fromthe analysis and make inference on key quantities of interest, both in a Bayesianand not Bayesian setting, and how to build a reference prior

This dissertation consists of four independent but related parts, each in a Chapter. The first part is an introductory. It serves as the background introduction and offer preparations for later parts. The second part discusses two population multivariate normal distributions with common covariance matrix. The goal for this part is to derive objective/non-informative priors for the parameterizations and use these priors to build up constructive random posteriors of the Kullback-Liebler (KL) divergence of the two multivariate normal populations, which is proportional to the distance between the two means, weighted by the common precision matrix. We use the Cholesky decomposition for re-parameterization of the precision matrix. The KL divergence is a true distance measurement for divergence between the two multivariate normal populations with common covariance matrix. Frequentist properties of the Bayesian procedure using these objective priors are studied through analytical and numerical tools. The third part considers the star-shape Gaussian graphical model, which is a special case of undirected Gaussian graphical models. It is a multivariate normal distribution where the variables are grouped into one "global" group of variable set and several "local" groups of variable set. When conditioned on the global variable set, the local variable sets are independent of each other. We adopt the Cholesky decomposition for re-parametrization of precision matrix and derive Jeffreys' prior, reference prior, and invariant priors for new parameterizations. The frequentist properties of the Bayesian procedure using these objective priors are also studied. The last part concentrates on the discussion of objective Bayesian analysis for partial correlation coefficient and its application to multivariate Gaussian models.
Ph. D.

The dissertation considers construction of confidence intervals for a cumulative distribution function F(z) and its inverse at some fixed points z and u on the basis of an i.i.d. sample where the sample size is relatively small. The sample is modeled as having the flexible Generalized Gamma distribution with all three parameters being unknown. This approach can be viewed as an alternative to nonparametric techniques which do not specify distribution of X and lead to less efficient procedures. The confidence intervals are constructed by objective Bayesian methods and use the Jeffreys noninformative prior. Performance of the resulting confidence intervals is studied via Monte Carlo simulations and compared to the performance of nonparametric confidence intervals based on binomial proportion. In addition, techniques for change point detection are analyzed and further evaluated via Monte Carlo simulations. The effect of a change point on the interval estimators is studied both analytically and via Monte Carlo simulations.
Ph.D.
Department of Mathematics
Arts and Sciences
Mathematics

Submitted by Luciana Sebin (lusebin@ufscar.br) on 2016-09-26T18:57:40Z No. of bitstreams: 1 DissACRS.pdf: 1390452 bytes, checksum: a5365fdbf745228c0174f2643b3f7267 (MD5)
Approved for entry into archive by Marina Freitas (marinapf@ufscar.br) on 2016-09-27T19:59:56Z (GMT) No. of bitstreams: 1 DissACRS.pdf: 1390452 bytes, checksum: a5365fdbf745228c0174f2643b3f7267 (MD5)
Approved for entry into archive by Marina Freitas (marinapf@ufscar.br) on 2016-09-27T20:00:01Z (GMT) No. of bitstreams: 1 DissACRS.pdf: 1390452 bytes, checksum: a5365fdbf745228c0174f2643b3f7267 (MD5)
Made available in DSpace on 2016-09-27T20:00:08Z (GMT). No. of bitstreams: 1 DissACRS.pdf: 1390452 bytes, checksum: a5365fdbf745228c0174f2643b3f7267 (MD5) Previous issue date: 2016-02-18
Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES)
In this work , we present an extension of the objective bayesian analysis made in Fonseca et al. (2008), based on Je reys priors for linear regression models with Student t errors, for which we consider the heteroscedasticity assumption. We show that the posterior distribution generated by the proposed Je reys prior, is proper. Through simulation study , we analyzed the frequentist properties of the bayesian estimators obtained. Then we tested the robustness of the model through disturbances in the response variable by comparing its performance with those obtained under another prior distributions proposed in the literature. Finally, a real data set is used to analyze the performance of the proposed model . We detected possible in uential points through the Kullback -Leibler divergence measure, and used the selection model criterias EAIC, EBIC, DIC and LPML in order to compare the models.
Neste trabalho, apresentamos uma extensão da análise bayesiana objetiva feita em Fonseca et al. (2008), baseada nas distribuicões a priori de Je reys para o modelo de regressão linear com erros t-Student, para os quais consideramos a suposicão de heteoscedasticidade. Mostramos que a distribuiçãoo a posteriori dos parâmetros do modelo regressão gerada pela distribuição a priori e própria. Através de um estudo de simulação, avaliamos as propriedades frequentistas dos estimadores bayesianos e comparamos os resultados com outras distribuições a priori encontradas na literatura. Além disso, uma análise de diagnóstico baseada na medida de divergência Kullback-Leiber e desenvolvida com analidade de estudar a robustez das estimativas na presença de observações atípicas. Finalmente, um conjunto de dados reais e utilizado para o ajuste do modelo proposto.

La reformulation du théorème de Bayes par R. de Cristofaro permet d'intégrer l'information sur le plan expérimental dans la loi a priori. En acceptant de transgresser les principes de vraisemblance et de la règle d'arrêt, un nouveau cadre théorique permet d'aborder le problème de la séquentialité dans l'inférence bayésienne. En considérant que l'information sur le plan expérimental est contenue dans l'information de Fisher, on dérive une famille de lois a priori à partir d'une vraisemblance directement associée à l'échantillonnage. Le cas de l'évaluation d'une proportion dans le contexte d'échantillonnages Binomiaux successifs conduit à considérer la loi Bêta-J. L'étude sur plusieurs plans séquentiels permet d'établir que l'"a priori de Jeffreys corrigé" compense le biais induit sur la proportion observée. Une application dans l'estimation ponctuelle montre le lien entre le paramétrage des lois Bêta-J et Bêta dans l'échantillonnage fixe. La moyenne et le mode des lois a posteriori obtenues présentent des propriétés fréquentistes remarquables. De même, l'intervalle de Jeffreys corrigé montre un taux de recouvrement optimal car la correction vient compenser l'effet de la règle d'arrêt sur les bornes. Enfin, une procédure de test, dont les erreurs s'interprètent à la fois en terme de probabilité bayésienne de l'hypothèse et de risques fréquentistes, est construite avec une règle d'arrêt et de rejet de H0 fondée sur une valeur limite du facteur de Bayes. On montre comment l'a priori de Jeffreys corrigé compense le rapport des évidences et garantit l'unicité des solutions, y compris lorsque l'hypothèse nulle est composite.

Submitted by Automa??o e Estat?stica (sst@bczm.ufrn.br) on 2017-01-13T12:23:37Z No. of bitstreams: 1 IsaacJalesCostaSouza_DISSERT.pdf: 808186 bytes, checksum: 0218f6e40a4dfea5b56a9d90f17e0bfb (MD5)
Approved for entry into archive by Arlan Eloi Leite Silva (eloihistoriador@yahoo.com.br) on 2017-01-23T13:11:35Z (GMT) No. of bitstreams: 1 IsaacJalesCostaSouza_DISSERT.pdf: 808186 bytes, checksum: 0218f6e40a4dfea5b56a9d90f17e0bfb (MD5)
Made available in DSpace on 2017-01-23T13:11:35Z (GMT). No. of bitstreams: 1 IsaacJalesCostaSouza_DISSERT.pdf: 808186 bytes, checksum: 0218f6e40a4dfea5b56a9d90f17e0bfb (MD5) Previous issue date: 2016-01-28
Neste trabalho ? apresentada uma abordagem bayesiana dos modelos pot?ncia normal bimodal (PNB) e pot?ncia normal bimodal assim?trico (PNBA). Primeiramente, apresentamos o modelo PNB e especificamos para este prioris n?o informativas e informativas do par?metroque concentra a bimodalidade (?). Em seguida, obtemos a distribui??o a posteriori pelo m?todo MCMC, o qual testamos a viabilidade de seu uso a partir de um diagn?stico de converg?ncia. Depois, utilizamos diferentes prioris informativas para ? e fizemos a an?lise de sensibilidadecom o intuito de avaliar o efeito da varia??o dos hiperpar?metros na distribui??o a posteriori. Tamb?m foi feita uma simula??o para avaliar o desempenho do estimador bayesiano utilizando prioris informativas. Constatamos que a estimativa da moda a posteriori apresentou em geralresultados melhores quanto ao erro quadratico m?dio (EQM) e vi?s percentual (VP) quando comparado ao estimador de m?xima verossimilhan?a. Uma aplica??o com dados bimodais reais foi realizada. Por ?ltimo, introduzimos o modelo de regress?o linear com res?duos PNB. Quanto ao modelo PNBA, tamb?m especificamos prioris informativas e n?o informativas para os par?metros de bimodalidade e assimetria. Fizemos o diagn?stico de converg?ncia para o m?todo MCMC, que tamb?m foi utilizado para obter a distribui??o a posteriori. Fizemos uma an?lise de sensibilidade, aplicamos dados reais no modelo e introduzimos o modelo de regress?o linear com res?duos PNBA.
In this paper it is presented a Bayesian approach to the bimodal power-normal (BPN) models and the bimodal asymmetric power-normal (BAPN). First, we present the BPN model, specifying its non-informative and informative parameter ? (bimodality). We obtain the posterior distribution by MCMC method, whose feasibility of use we tested from a convergence diagnose. After that, We use different informative priors for ? and we do a sensitivity analysis in order to evaluate the effect of hyperparameters variation on the posterior distribution. Also, it is performed a simulation to evaluate the performance of the Bayesian estimator using informative priors. We noted that the Bayesian method shows more satisfactory results when compared to the maximum likelihood method. It is performed an application with bimodal data. Finally, we introduce the linear regression model with BPN error. As for the BAPN model we also specify informative and uninformative priors for bimodality and asymmetry parameters. We do the MCMC Convergence Diagnostics, which is also used to obtain the posterior distribution. We do a sensitivity analysis, applying actual data in the model and we introducing the linear regression model with PNBA error.

A literatura referente a testes de hipótese em modelos auto-regressivos que apresentam uma possível raiz unitária é bastante vasta e engloba pesquisas oriundas de diversas áreas. Nesta dissertação, inicialmente, buscou-se realizar uma revisão dos principais resultados existentes, oriundos tanto da visão clássica quanto da bayesiana de inferência. No que concerne ao ferramental clássico, o papel do movimento browniano foi apresentado de forma detalhada, buscando-se enfatizar a sua aplicabilidade na dedução de estatísticas assintóticas para a realização dos testes de hipótese relativos à presença de uma raíz unitária. Com relação à inferência bayesiana, foi inicialmente conduzido um exame detalhado do status corrente da literatura. A seguir, foi realizado um estudo comparativo em que se testa a hipótese de raiz unitária com base na probabilidade da densidade a posteriori do parâmetro do modelo, considerando as seguintes densidades a priori: Flat, Jeffreys, Normal e Beta. A inferência foi realizada com base no algoritmo Metropolis-Hastings, usando a técnica de simulação de Monte Carlo por Cadeias de Markov (MCMC). Poder, tamanho e confiança dos testes apresentados foram computados com o uso de séries simuladas. Finalmente, foi proposto um critério bayesiano de seleção de modelos, utilizando as mesmas distribuições a priori do teste de hipótese. Ambos os procedimentos foram ilustrados com aplicações empíricas à séries temporais macroeconômicas.
Testing for unit root hypothesis in non stationary autoregressive models has been a research topic disseminated along many academic areas. As a first step for approaching this issue, this dissertation includes an extensive review highlighting the main results provided by Classical and Bayesian inferences methods. Concerning Classical approach, the role of brownian motion is discussed in a very detailed way, clearly emphasizing its application for obtaining good asymptotic statistics when we are testing for the existence of a unit root in a time series. Alternatively, for Bayesian approach, a detailed discussion is also introduced in the main text. Then, exploring an empirical façade of this dissertation, we implemented a comparative study for testing unit root based on a posteriori model's parameter density probability, taking into account the following a priori densities: Flat, Jeffreys, Normal and Beta. The inference is based on the Metropolis-Hastings algorithm and on the Monte Carlo Markov Chains (MCMC) technique. Simulated time series are used for calculating size, power and confidence intervals for the developed unit root hypothesis test. Finally, we proposed a Bayesian criterion for selecting models based on the same a priori distributions used for developing the same hypothesis tests. Obviously, both procedures are empirically illustrated through application to macroeconomic time series.

Research Doctorate - Doctor of Philosophy (PhD)
Interval estimation of the Binomial parameter è, representing the true probability of a success, is a problem of long standing in statistical inference. The landmark work is by Bayes (1763) who applied the uniform prior to derive the Beta posterior that is the normalised Binomial likelihood function. It is not well known that Bayes favoured this ‘noninformative’ prior as a result of considering the observable random variable x as opposed to the unknown parameter è, which is an important difference. In this thesis we develop additional arguments in favour of the uniform prior for estimation of è. We start by describing the frequentist and Bayesian approaches to interval estimation. It is well known that for common continuous models, while different in interpretation, frequentist and Bayesian intervals are often identical, which is directly related to the existence of a pivotal quantity. The Binomial model, and its Poisson sister also, lack a pivotal quantity, despite having sufficient statistics. Lack of a pivotal quantity is the reason why there is no consensus on one particular estimation method, more so than its discreteness: frequentist (unconditional) coverage depends on è. Exact methods guarantee minimum coverage to be at least equal to nominal and approximate methods aim for mean coverage to be close to nominal. We agree with what seems like the majority of frequentists, that exact methods are too conservative in practice, and show additional undesirable properties. This includes more recent ‘short’ exact intervals. We argue that Bayesian intervals based on noninformative priors are preferable to the family of frequentist approximate intervals, some of which are wider than exact intervals for particular data values. A particular property of the interval based on the uniform prior is that its mean coverage is exactly equal to nominal. However, once committed to the Bayesian approach there is no denying that the current preferred choice, by ‘objective’ Bayesians, is the U-shaped Jeffreys prior which results from various methods aimed at finding noninformative priors. The most successful such method seems to be reference analysis which has led to sensible priors in previously unsolved problems, concerning multiparameter models that include ‘nuisance’ parameters. However, we argue that there is a class of models for which the Jeffreys/reference prior may be suboptimal and that in the case of the Binomial distribution the requirement of a uniform prior predictive distribution leads to a more reasonable ‘consensus’ prior.

Research Doctorate - Doctor of Philosophy (PhD)
Interval estimation of the Binomial parameter è, representing the true probability of a success, is a problem of long standing in statistical inference. The landmark work is by Bayes (1763) who applied the uniform prior to derive the Beta posterior that is the normalised Binomial likelihood function. It is not well known that Bayes favoured this ‘noninformative’ prior as a result of considering the observable random variable x as opposed to the unknown parameter è, which is an important difference. In this thesis we develop additional arguments in favour of the uniform prior for estimation of è. We start by describing the frequentist and Bayesian approaches to interval estimation. It is well known that for common continuous models, while different in interpretation, frequentist and Bayesian intervals are often identical, which is directly related to the existence of a pivotal quantity. The Binomial model, and its Poisson sister also, lack a pivotal quantity, despite having sufficient statistics. Lack of a pivotal quantity is the reason why there is no consensus on one particular estimation method, more so than its discreteness: frequentist (unconditional) coverage depends on è. Exact methods guarantee minimum coverage to be at least equal to nominal and approximate methods aim for mean coverage to be close to nominal. We agree with what seems like the majority of frequentists, that exact methods are too conservative in practice, and show additional undesirable properties. This includes more recent ‘short’ exact intervals. We argue that Bayesian intervals based on noninformative priors are preferable to the family of frequentist approximate intervals, some of which are wider than exact intervals for particular data values. A particular property of the interval based on the uniform prior is that its mean coverage is exactly equal to nominal. However, once committed to the Bayesian approach there is no denying that the current preferred choice, by ‘objective’ Bayesians, is the U-shaped Jeffreys prior which results from various methods aimed at finding noninformative priors. The most successful such method seems to be reference analysis which has led to sensible priors in previously unsolved problems, concerning multiparameter models that include ‘nuisance’ parameters. However, we argue that there is a class of models for which the Jeffreys/reference prior may be suboptimal and that in the case of the Binomial distribution the requirement of a uniform prior predictive distribution leads to a more reasonable ‘consensus’ prior.