The aim of this work is to present an inference and diagnostic study of an extension of the lifetime distribution family proposed by Birnbaum and Saunders (1969a,b). This extension is obtained by considering a skew-elliptical distribution instead of the normal distribution. Specifically, in this work we develop a Birnbaum-Saunders (BS) distribution type based on scale mixtures of skew-normal distributions (SMSN). The resulting family of lifetime distributions represents a robust extension of the usual BS distribution. Based on this family, we reproduce the usual properties of the BS distribution, and present an estimation method based on the EM algorithm. In addition, we present regression models associated with the BS distributions (based on scale mixtures of skew-normal), which are developed as an extension of the sinh-normal distribution (Rieck and Nedelman, 1991). For this model we consider an estimation and diagnostic study for uncensored data. / O objetivo deste trabalho é apresentar um estudo de inferência e diagnóstico em uma extensão da família de distribuições de tempos de vida proposta por Birnbaum e Saunders (1969a,b). Esta extensão é obtida ao considerar uma distribuição skew-elíptica em lugar da distribuição normal. Especificamente, neste trabalho desenvolveremos um tipo de distribuição Birnbaum-Saunders (BS) baseda nas distribuições mistura de escala skew-normal (MESN). Esta família resultante de distribuições de tempos de vida representa uma extensão robusta da distribuição BS usual. Baseado nesta família, vamos reproduzir as propriedades usuais da distribuição BS, e apresentar um método de estimação baseado no algoritmo EM. Além disso, vamos apresentar modelos de regressão associado à distribuições BS (baseada na distribuição mistura de escala skew-normal), que é desenvolvida como uma extensão da distribuição senh-normal (Rieck e Nedelman, 1991), para estes vamos considerar um estudo de estimação e diagnóstisco para dados sem censura.

In 2007, Domínguez-Molina et al. obtained the moment generating function (mgf) of the matrix variate closed skew normal distribution. In this paper, we use their mgf to obtain the first two moments and some additional properties of quadratic forms for the matrix variate skew normal distributions. The quadratic forms are particularly interesting because they are essentially correlation tests that introduce a new type of orthogonality condition.

No description available.

My dissertation mainly focuses on the regularization of parameters in the multivariate linear regression under different assumptions on the distribution of the errors. It consists of two topics where we develop iterative procedures to construct sparse estimators for both the regression coefficient and scale matrices simultaneously, and a third topic where we develop a method for testing if the skewness parameter in the skew-normal distribution is parallel to one of the eigenvectors of the scale matrix. In the first project, we propose a robust procedure for constructing a sparse estimator of a multivariate regression coefficient matrix that accounts for the correlations of the response variables. Robustness to outliers is achieved using heavy-tailed t distributions for the multivariate response, and shrinkage is introduced by adding to the negative log-likelihood l1 penalties on the entries of both the regression coefficient matrix and the precision matrix of the responses. Taking advantage of the hierarchical representation of a multivariate t distribution as the scale mixture of normal distributions and the EM algorithm, the optimization problem is solved iteratively where at each EM iteration suitably modified multivariate regression with covariance estimation (MRCE) algorithms proposed by Rothman, Levina and Zhu are used. We propose two new optimization algorithms for the penalized likelihood, called MRCEI and MRCEII, which differ from MRCE in the way that the tuning parameters for the two matrices are selected. Estimating the degrees of freedom when penalizing the entries of the matrices presents new computational challenges. A simulation study and real data analysis demonstrate that the MRCEII, which selects the tuning parameter of the precision matrix of the multiple responses using the Cp criterion, generally does the best among all methods considered in terms of the prediction error, and MRCEI outperforms the MRCE methods when the regression coefficient matrix is less sparse. The second project is motivated by the existence of the skewness in the data for which the symmetric distribution assumption on the errors does not hold. We extend the procedure we have proposed to the case where the errors in the multivariate linear regression follow a multivariate skew-normal or skew-t distribution. Based on the convenient representation of skew-normal and skew-t as well as the EM algorithm, we develop an optimization algorithm, called MRST, to iteratively minimize the negative penalized log-likelihood. We also carry out a simulation study to assess the performance of the method and illustrate its application with one real data example. In the third project, we discuss the asymptotic distributions of the eigenvalues and eigenvectors for the MLE of the scale matrix in a multivariate skew-normal distribution. We propose a statistic for testing whether the skewness vector is proportional to one of the eigenvectors of the scale matrix based on the likelihood ratio. Under the alternative, the likelihood is maximized numerically with two different ways of parametrization for the scale matrix: Modified Cholesky Decomposition (MCD) and Givens Angle. We conduct a simulation study and show that the statistic obtained using Givens Angle parametrization performs well and is more reliable than that obtained using MCD.

This research explores factor analysis applied to data from skewed distributions for the general skew model, the selection-elliptical model, the selection-normal model, the skew-elliptical model and the skew-normal model for finite sample sizes. In terms of asymptotics, or large sample sizes, quasi-maximum likelihood methods are broached numerically. The skewed models are formed using selection distribution theory, which is based on Rao’s weighted distribution theory. The models assume the observed variable of the factor model is from a skewed distribution by defining the distribution of the unobserved common factors skewed and the unobserved unique factors symmetric. Numerical examples are provided using maximum likelihood selection skew-normal factor analysis. The numerical examples, such as maximum likelihood parameter estimation with the resolution of the “sign switching” problem and model fitting using likelihood methods, illustrate that the selection skew-normal factor analysis model better fits skew-normal data than does the normal factor analysis model.

Matrix variate skew elliptically contoured distributions generalize several classes of important distributions. This paper defines and explores matrix variate skew elliptically contoured distributions. In particular, we discuss two stochastic representations of the matrix variate skew elliptically contoured distributions.

This thesis explores the problems of fitting antedependence (AD) models and partial antecorrelation (PAC) models to continuous non-Gaussian longitudinal data. AD models impose certain conditional independence relations among the measurements within each subject, while PAC models characterize the partial correlation relations. The models are parsimonious and useful for data exhibiting time-dependent correlations. Since the relation of conditional independence among variables is rather restrictive, we first consider an autoregressively characterized PAC model with independent asymmetric Laplace (ALD) innovations and prove that this model is an AD model. The ALD distribution previously has been applied to quantile regression and has shown promise for modeling asymmetrically distributed ecological data. In addition, the double exponential distribution, a special case of the ALD, has played an important role in fitting symmetric finance and hydrology data. We give the distribution of a linear combination of independent standard ALD variables in order to derive marginal distributions for the model. For the model estimation problem, we propose an iterative algorithm for the maximum likelihood estimation. The estimation accuracy is illustrated by some numerical examples as well as some longitudinal data sets. The second component of this dissertation focuses on AD multivariate skew normal models. The multivariate skew normal distribution not only shares some nice properties with multivariate normal distributions but also allows for any value of skewness. We derive necessary and sufficient conditions on the shape and covariance parameters for multivariate skew normal variables to be AD(p) for some p. Likelihood-based estimation for balanced and monotone missing data as well as likelihood ratio hypothesis tests for the order of antedependence and for zero skewness under the models are presented. Since the class of skew normal random variables is closed under the addition of independent standard normal random variables, we then consider an autoregressively characterized PAC model with a combination of independent skew normal and normal innovations. Explicit expressions for the marginals, which all have skew normal distributions, and maximum likelihood estimates of model parameters, are given. Numerical results show that these three proposed models may provide reasonable fits to some continuous non-Gaussian longitudinal data sets. Furthermore, we compare the fits of these models to the Treatment A cattle growth data using penalized likelihood criteria, and demonstrate that the AD(2) multivariate skew normal model fits the data best among those proposed models.

Orientadores: Caio Lucidius Naberezny Azevedo, Dalton Francisco de Andrade / Dissertação (mestrado) - Universidade Estadual de Campinas, Instituto de Matemática Estatística e Computação Científica / Made available in DSpace on 2018-08-24T22:30:44Z (GMT). No. of bitstreams: 1 PadillaGomez_JuanLeonardo_M.pdf: 10775900 bytes, checksum: 50bc9965f728b4b04b42b7428c3ec8ab (MD5) Previous issue date: 2014 / Resumo: No presente trabalho propõe-se novos modelos da Teoria de Resposta ao Item Multidimensional (TRIM) para respostas dicotômicas ou dicotomizadas considerando uma estrutura de grupos múltiplos. Para as distribuições dos traços latentes de cada grupo, propõe-se uma nova parametrização da distribuição normal assimétrica multivariada centrada, que combina as propostas de Lachos (2004) e de Arellano-Valle et.al (2008), a qual não só garante a identificabilidade dos modelos aqui introduzidos, mas também facilita a interpretação e estimação dos seus parâmetros. Portanto, nosso modelo representa uma alternativa interessante, para solucionar os problemas de falta de identificabilidade encontrados por Matos (2010) e Nojosa (2008), nos modelos multidimensionais assimétricos de um único grupo por eles desenvolvidos. Estudos de simulação, considerando vários cenários de interesse prático, foram conduzidos a fim de avaliar o potencial da tríade: modelagem, métodos de estimação e ferramentas de diagnósticos. Os resultados indicam que os modelos considerando a assimetria nos traços latentes, em geral, forneceram estimativas mais acuradas que os modelos tradicionais. Para a seleção de modelos, utilizou-se o critério de informação deviance (DIC), os valores esperados do critério de informação de Akaike (EAIC) e o critério de informação bayesiano (EBIC). Em relação à verificação da qualidade do ajuste de modelos, explorou-se alguns métodos de checagem preditiva a posteriori, os quais fornecem meios para avaliar a qualidade tanto do instrumento de medida, quanto o ajuste do modelo de um ponto de vista global e em relação à suposições específicas, entre elas a dimensão do teste. Com relação aos métodos de estimação, adaptou-se e implementou-se vários algoritmos MCMC propostos na literatura para outros modelos, inclusive a proposta de aceleração de convergência de González (2004), os quais foram comparados em relação aos aspectos de qualidade de convergência através do critério de tamanho efetivo da amostra de Sahu (2002). A análise de um conjunto de dados reais, referente à primeira fase do vestibular da UNICAMP de 2013 também foi realizada / Abstract: In this work it is proposed a new class of Multidimensional Item Response Theory (MIRT) models for dichotomic or dichotomized answers considering a multiple group structure. For the latent traits distribution of each group, it is proposed a new parametrization of the centered multivariate skew normal distribution, which combines the proposed by Lachos (2004) and the one proposed by Arellano-Valle et.al (2008), which not only ensures de identifiability of our proposed models, but also it makes simpler the interpretation and estimation of their parameters. Hence, our model stands as an important alternative, in order to solve the identifiability problems found for the one group multidimensional skewed models proposed by Matos (2010) and Nojosa (2008). Simulation studies, taking into account some situations of practical interest, were conducted in order to evaluate the potential of the triad: modeling, estimation methods and diagnostic tools. The results indicate that the models considering a skew component on the latent traits, in general, produced more accurate results than those ones obtained with the symmetric models. For model selection, it was used the deviance information criterion (DIC), the expected values of both the Akaike¿s information criterion (EAIC) and bayesian information criteron (EBIC). Concerning assessment of model fit quality, it was explored posterior predictive checking methods, which allows for evaluating the quality of the measure instrument as well as the quality fit of the model from a global point of view and related to specific assumptions, as the test dimensionality. Concerning the estimation methods, it was adopted and implemented several MCMC algorithms proposed in the literature for other models, including the convergence accelerating propose algorithm by Gonzalez (2004), which were compared concerning some convergence quality aspects through the Sahu (2002) effective sample size. The analysis of a real data set, from the 2013 first stage of the UNICAMP admission exam was done as well / Mestrado / Estatistica / Mestre em Estatística

Orientador: Caio Lucidius Naberezny Azevedo / Dissertação (mestrado) - Universidade Estadual de Campinas, Instituto de Matemática, Estatística e Computação Científica / Made available in DSpace on 2018-08-20T09:23:25Z (GMT). No. of bitstreams: 1 Santos_JoseRobertoSilvados_M.pdf: 2068782 bytes, checksum: f8dc91d2f7f6091813ba229dc12991f4 (MD5) Previous issue date: 2012 / Resumo: Uma das suposições dominantes nos modelos de resposta ao item (MRI) é a suposição de normalidade simétrica para modelar o comportamento dos traços latentes. No entanto, tal suposição tem sido questionada em vários trabalhos como, por exemplo, nos trabalhos de Micceri (1989) e Bazán et.al (2006). Recentemente Azevedo et.al (2011) propuseram um MRI com distribuição normal assimétrica centralizada para os traços latentes, considerando a estrutura de um único grupo de indivíduos. No presente trabalho fazemos uma extensão desse modelo para o caso de grupos múltiplos. Desenvolvemos dois algoritmos MCMC para estimação dos parâmetros utilizando a estrutura de dados aumentados para representar a função de resposta ao item (FRI), veja Albert (1992). O primeiro é um amostrador de Gibbs com passos de Metropolis-Hastings. No segundo utilizamos representações estocásticas (gerando uma estrutura hierárquica) das densidades a priori dos traços latentes e parâmetros populacionais conseguindo, assim, formas conhecidas para todas as distribuições condicionais completas, o que nos possibilitou desenvolver o amostrador de Gibbs completo. Comparamos esses algoritmos utilizando como critério o tamanho efetivo de amostra, veja Sahu (2002). O amostrador de Gibbs completo obteve o melhor desempenho. Também avaliamos o impacto do número de respondentes por grupo, número de itens por grupo, número de itens comuns, assimetria da distribuição do grupo de referência e priori, na recuperação dos parâmetros. Os resultados indicaram que nosso modelo recuperou bem todos os parâmetros, principalmente, quando utilizamos a priori de Jeffreys. Além disso, o número de itens por grupo e o número de examinados por grupo, mostraram ter um alto impacto na recuperação dos traços latentes e parâmetros dos itens, respectivamente. Analisamos um conjunto de dados reais que apresenta indícios de assimetria na distribuição dos traços latentes de alguns grupos. Os resultados obtidos com o nosso modelo confirmam a presença de assimetria na maioria dos grupos. Estudamos algumas medidas de diagnóstico baseadas na distribuição preditiva de medidas de discrepância adequadas. Por último, comparamos os modelos simétrico e assimétrico utilizando os critérios sugeridos por Spiegelhalter et al. (2002). O modelo assimétrico se ajustou melhor aos dados segundo todos os critérios / Abstract: An usual assumption for parameter estimation in the Item Response Models (IRM) is to assume that the latent traits are random variables which follow a normal distribution. However, many works suggest that this assumption does not apply in many cases. For example, the works of Micceri (1989) and Bazán (2006). Recently Azevedo et.al (2011) proposed an IRM with skew-normal distribution under the centred parametrization for the latent traits, considering one single group of examinees. In the present work, we developed an extension of this model to account for multiple groups. We developed two MCMC algorithms to parameter estimation using the augmented data structure to represent the Item response function (IRF), see Albert (1992). The First is a Metropolis-Hastings within Gibbs sampling. In the second, we use stochastic representations (creating a hierarchical structure) in the prior distribution of the latent traits and population parameters. Therefore, we obtained known full conditional distributions, which enabled us to develop the full Gibbs sampler. We compared these algorithms using the effective sample size criteria, see Sahu (2002). The full Gibbs sampling presented the best performance. We also evaluated the impact of the number of examinees per group, number of items per group, number of common items, priors and asymmetry of the reference group, on the parameter recovery. The results indicated that our approach recovers properly all parameters, mainly, when we consider the Jeffreys prior. Furthermore, the number of items per group and the number of examinees per group, showed to have a high impact on the recovery of the true of latent traits and item parameters, respectively. We analyze a real data set in which we found an evidence of asymmetry in the distribution of latent traits of some groups. The results obtained with our model confirmed the presence of asymmetry in most groups. We studied some diagnostic measures based on predictive distribution of appropriate discrepancy measures. Finally, we compared the symmetric and asymmetric models using the criteria suggested by Spiegelhalter et al. (2002). The asymmetrical model fits better according to all criteria / Mestrado / Estatistica / Mestre em Estatística

Dans le contexte de la théorie bayésienne et de théorie de la décision, l'estimation d'une densité prédictive d'une variable aléatoire occupe une place importante. Typiquement, dans un cadre paramétrique, il y a présence d’information additionnelle pouvant être interprétée sous forme d’une contrainte. Cette thèse porte sur des stratégies et des améliorations, tenant compte de l’information additionnelle, pour obtenir des densités prédictives efficaces et parfois plus performantes que d’autres données dans la littérature. Les résultats s’appliquent pour des modèles avec données gaussiennes avec ou sans une variance connue. Nous décrivons des densités prédictives bayésiennes pour les coûts Kullback-Leibler, Hellinger, Kullback-Leibler inversé, ainsi que pour des coûts du type $\alpha-$divergence et établissons des liens avec les familles de lois de probabilité du type \textit{skew--normal}. Nous obtenons des résultats de dominance faisant intervenir plusieurs techniques, dont l’expansion de la variance, les fonctions de coût duaux en estimation ponctuelle, l’estimation sous contraintes et l’estimation de Stein. Enfin, nous obtenons un résultat général pour l’estimation bayésienne d’un rapport de deux densités provenant de familles exponentielles. / Abstract: In the context of Bayesian theory and decision theory, the estimation of a predictive density of a random variable represents an important and challenging problem. Typically, in a parametric framework, usually there exists some additional information that can be interpreted as constraints. This thesis deals with strategies and improvements that take into account the additional information, in order to obtain effective and sometimes better performing predictive densities than others in the literature. The results apply to normal models with a known or unknown variance. We describe Bayesian predictive densities for Kullback--Leibler, Hellinger, reverse Kullback-Leibler losses as well as for α--divergence losses and establish links with skew--normal densities. We obtain dominance results using several techniques, including expansion of variance, dual loss functions in point estimation, restricted parameter space estimation, and Stein estimation. Finally, we obtain a general result for the Bayesian estimator of a ratio of two exponential family densities.