The recent popularity of Bayesian inference is due to the practical advantages of the Bayesian approach. The Bayesian analysis makes it possible to reflect ones prior beliefs into the analysis. In this thesis, we explore some asymptotic results in Bayesian nonparametric inference for restricted and unrestricted space of distributions. This thesis is divided into two parts. In the first part, we employ the Dirichlet process in a hypothesis testing framework to propose a Bayesian nonparametric chi-squared goodness-of-fit test. Our suggested method corresponds to Lo's Bayesian bootstrap procedure for chi-squared goodness-of-fit test. Indeed, our bootstrap rectifies some shortcomings of regular bootstrap which only counts number of observations falling in each bin in contingency tables. We consider the Dirichlet process as the prior for the distribution of data and carry out the test based on the Kullback-Leibler distance between the Dirichlet process posterior and the hypothesized distribution. We prove that this distance asymptotically converges to the same chi-squared distribution as the classical frequentist's chi-squared test. Moreover, the results are generalized to the chi-squared test of independence for contingency tables. In the second part, our main focus is on Bayesian nonparametric inference for a restricted group of distributions called spherically symmetric distributions. We describe a Bayesian nonparametric approach to perform an inference for a bivariate spherically symmetric distribution. We place a Dirichlet invariant process prior on the set of all bivariate spherically symmetric distributions and derive the Dirichlet invariant process posterior. Indeed, our approach is an extension of the Dirichlet invariant process for the symmetric distributions on the real line to bivariate spherically symmetric distribution where the underlying distribution is invariant under a finite group of rotations. Further, we obtain the Dirichlet invariant process posterior for the infinite transformation group and we prove that it approaches a certain Dirichlet process. Finally, we develop our approach to obtain the Bayesian nonparametric posterior distribution for functionals of the distribution's support when the support satisfies certain symmetry conditions. When symmetry holds with respect to the parallel lines of axes (for example, in two dimensional space x = a and y = b) we employ our approach to approximate the distribution of certain functionals such as area and perimeter for the support of the distribution. This suggests a Bayesian nonparametric bootstrapping scheme. The estimates can be derived based on posterior averaging. Then, our simulation results demonstrate that our suggested bootstrapping technique improves the accuracy of the estimates.

We present an algorithm for imputation of incomplete datasets based on Bayesian exchangeability through Pólya sampling. Each (donee) unit with a missing value is imputed multiple times by observed (real) values on units from a donor pool. The donor pools are constructed using auxiliary variables. Several features from kernel estimation are used to counteract unbalances that are due to sparse and bounded data. Three balancing features can be used with only one single continuous auxiliary variable, but an additional fourth feature need, multiple continuous auxiliary variables. They mainly contribute by reducing nonresponse bias. We examine how the donor pool size should be determined, that is the number of potential donors within the pool. External information is shown to be easily incorporated in the imputation algorithm. Our simulation studies show that with a study variable which can be seen as a function of one or two continuous auxiliaries plus residual noise, the method performs as well or almost as well as competing methods when the function is linear, but usually much better when the function is nonlinear. / <p>At the time of the doctoral defense, the following papers were unpublished and had a status as follows: Paper 1: In press. Paper 3: Submitted. Paper 4: Submitted.</p>

Experimentos nos quais uma ou mais variáveis respostas são influênciadas por diversos fatores quantitativos são bastante comuns nas áreas agrícola, química, biológica, dentre outras. Nesse caso, o problema de pesquisa consiste em se estudar essa relação, sendo de grande utilidade o uso da metodologia de superfícies de resposta (MSR). Nesse contexto, a determinação dos níveis dos fatores que otimizam a resposta consiste inicialmente na obtenção das coordenadas do ponto estacionário do modelo ajustado. No entanto, como o modelo verdadeiro é desconhecido, é interessante obter uma região de confiança das coordenadas verdadeiras de modo a avaliar a precisão da estimativa obtida. Foram abordados aqui os procedimentos para construção de regiões de confiança para as coordenadas do ponto estacionário em diferentes situações considerando-se a forma das superfícies analisadas e a distribuição e magnitude da variância dos erros do modelo. Foram utilizadas a metodologia de Box e Hunter (1954) (BH), "bootstrap" e "bootstrap" Bayesiano aliados ao cálculo da distância de Mahalanobis entre as coordenadas do ponto estacionários da amostra observada e aquelas obtidas por meio das estimativas "bootstrap"(BM e BBM), e métodos "bootstrap" e "bootstrap" Bayesiano aliados a métodos não paramétricos de estimação de funções densidade de probabilidade (BNP e BBNP). A avaliaçãoda metodologia foi realizada por meio de simulação e foi aplicada a um conjunto de dados de produção de amendoim. No estudo de simulação, a metodologia BH, baseada na distribuição normal dos erros, apresentou um bom desempenho em todas as situações analisadas, havendo concordância entre as regiões de confiança nominais e reais, mesmo naquelas em que essa distribuição é bastante assimétrica. Este mesmo comportamento foi observado para os métodos BM e BBM. No entanto, os métodos BNP e BBNP não apresentaram um desempenho satisfatório, resultando em um nível de significância real menor que o nominal para os autovalores com menor valor absoluto, gerando regiões de confiança maiores. No caso de autovalores com maior valor absoluto observou-se situação inversa. No caso da análise do conjunto de dados de amendoim os métodos BH, BM e BNP apresentaram regiões de confiança mais amplas comparativamente aos métodos BBM e BBNP. No entanto, os valores das estimativas do "bootstrap" Bayesiano são mais próximas das estimativas de mínimos quadrados e apresentam menor dispersão o que explica a menor área da região de confiança. / Experiments in which one or more response variables are influenced by several quantitative factors are very common in agricultural, chemistry, biology and other areas. In this case, the research question consists in studying this relation, being of great utility the use of response surface methodology (RSM). In this context determining the level of the factors that optimize the response consists finding the coordinates of the stationary point of the model. However, as the true model is unknown, it is of interest to obtain a confidence region of the true coordinates to analyze the precision of the obtained estimate. The procedures for the construction of confidence regions for the coordinates of the stationary point were studied in diferent situations, considering the shape of the surfaces analyzed and the distribution and magnitude of the variance errors. The methodology of Box and Hunter (1954) (BH), bootstrap and Bayesian bootstrap with Mahalanobis distance among the coordinates of the stationary point of the observed sample and those obtained using bootstrap estimates(BM and BBM) and bootstrap and Bayesian bootstrap with non-parametric methods for density estimation (BNP and BBNP) were compared. The methodology evaluation was realized by means of simulation and applied to a peanut yields data set. In simulation study the BH methodology, which is based in normal distribution of errors, presented a good performance in all of the analyzed situations, having concordance among the nominal and real confidence regions, even in those which this distribution is fairly asymmetric. This behavior was also observed in BM and BBM methods. The BNP and BBNP methods did not presented a satisfactory performance, resulting in a real significance level lower than the nominal for the eigenvalue with lower absolute value, generating bigger confidence regions. The inverse was observed using eigenvalue with higher absolute value. In the analysis of the peanut yields data set the BH, BM and BNP methods presented confidence regions larger than the BBM and BBNP methods. The Bayesian bootstrap estimate values are closer of the minimum square estimates and present less dispersion what explain the confidence region lower area.

Experimentos nos quais uma ou mais variáveis respostas são influênciadas por diversos fatores quantitativos são bastante comuns nas áreas agrícola, química, biológica, dentre outras. Nesse caso, o problema de pesquisa consiste em se estudar essa relação, sendo de grande utilidade o uso da metodologia de superfícies de resposta (MSR). Nesse contexto, a determinação dos níveis dos fatores que otimizam a resposta consiste inicialmente na obtenção das coordenadas do ponto estacionário do modelo ajustado. No entanto, como o modelo verdadeiro é desconhecido, é interessante obter uma região de confiança das coordenadas verdadeiras de modo a avaliar a precisão da estimativa obtida. Foram abordados aqui os procedimentos para construção de regiões de confiança para as coordenadas do ponto estacionário em diferentes situações considerando-se a forma das superfícies analisadas e a distribuição e magnitude da variância dos erros do modelo. Foram utilizadas a metodologia de Box e Hunter (1954) (BH), "bootstrap" e "bootstrap" Bayesiano aliados ao cálculo da distância de Mahalanobis entre as coordenadas do ponto estacionários da amostra observada e aquelas obtidas por meio das estimativas "bootstrap"(BM e BBM), e métodos "bootstrap" e "bootstrap" Bayesiano aliados a métodos não paramétricos de estimação de funções densidade de probabilidade (BNP e BBNP). A avaliaçãoda metodologia foi realizada por meio de simulação e foi aplicada a um conjunto de dados de produção de amendoim. No estudo de simulação, a metodologia BH, baseada na distribuição normal dos erros, apresentou um bom desempenho em todas as situações analisadas, havendo concordância entre as regiões de confiança nominais e reais, mesmo naquelas em que essa distribuição é bastante assimétrica. Este mesmo comportamento foi observado para os métodos BM e BBM. No entanto, os métodos BNP e BBNP não apresentaram um desempenho satisfatório, resultando em um nível de significância real menor que o nominal para os autovalores com menor valor absoluto, gerando regiões de confiança maiores. No caso de autovalores com maior valor absoluto observou-se situação inversa. No caso da análise do conjunto de dados de amendoim os métodos BH, BM e BNP apresentaram regiões de confiança mais amplas comparativamente aos métodos BBM e BBNP. No entanto, os valores das estimativas do "bootstrap" Bayesiano são mais próximas das estimativas de mínimos quadrados e apresentam menor dispersão o que explica a menor área da região de confiança. / Experiments in which one or more response variables are influenced by several quantitative factors are very common in agricultural, chemistry, biology and other areas. In this case, the research question consists in studying this relation, being of great utility the use of response surface methodology (RSM). In this context determining the level of the factors that optimize the response consists finding the coordinates of the stationary point of the model. However, as the true model is unknown, it is of interest to obtain a confidence region of the true coordinates to analyze the precision of the obtained estimate. The procedures for the construction of confidence regions for the coordinates of the stationary point were studied in diferent situations, considering the shape of the surfaces analyzed and the distribution and magnitude of the variance errors. The methodology of Box and Hunter (1954) (BH), bootstrap and Bayesian bootstrap with Mahalanobis distance among the coordinates of the stationary point of the observed sample and those obtained using bootstrap estimates(BM and BBM) and bootstrap and Bayesian bootstrap with non-parametric methods for density estimation (BNP and BBNP) were compared. The methodology evaluation was realized by means of simulation and applied to a peanut yields data set. In simulation study the BH methodology, which is based in normal distribution of errors, presented a good performance in all of the analyzed situations, having concordance among the nominal and real confidence regions, even in those which this distribution is fairly asymmetric. This behavior was also observed in BM and BBM methods. The BNP and BBNP methods did not presented a satisfactory performance, resulting in a real significance level lower than the nominal for the eigenvalue with lower absolute value, generating bigger confidence regions. The inverse was observed using eigenvalue with higher absolute value. In the analysis of the peanut yields data set the BH, BM and BNP methods presented confidence regions larger than the BBM and BBNP methods. The Bayesian bootstrap estimate values are closer of the minimum square estimates and present less dispersion what explain the confidence region lower area.

Le traitement des images satellitaires est considéré comme l'un des domaines les plus intéressants dans les domaines de traitement d'images numériques. Les images satellitaires peuvent être dégradées pour plusieurs raisons, notamment les mouvements des satellites, les conditions météorologiques, la dispersion et d'autres facteurs. Plusieurs méthodes d'amélioration et de restauration des images satellitaires ont été étudiées et développées dans la littérature. Les travaux présentés dans cette thèse se concentrent sur la restauration des images satellitaires par des techniques de filtrage statistique non linéaire. Dans un premier temps, nous avons proposé une nouvelle méthode pour restaurer les images satellitaires en combinant les techniques de restauration aveugle et non aveugle. La raison de cette combinaison est d'exploiter les avantages de chaque technique utilisée. Dans un deuxième temps, de nouveaux algorithmes statistiques de restauration d'images basés sur les filtres non linéaires et l'estimation non paramétrique de densité multivariée ont été proposés. L'estimation non paramétrique de la densité à postériori est utilisée dans l'étape de ré-échantillonnage du filtre Bayésien bootstrap pour résoudre le problème de la perte de diversité dans le système de particules. Enfin, nous avons introduit une nouvelle méthode de la combinaison hybride pour la restauration des images basée sur la transformée en ondelettes discrète (TOD) et les algorithmes proposés à l'étape deux, et nos avons prouvé que les performances de la méthode combinée sont meilleures que les performances de l'approche TOD pour la réduction du bruit dans les images satellitaires dégradées. / Satellite image processing is considered one of the more interesting areas in the fields of digital image processing. Satellite images are subject to be degraded due to several reasons, satellite movements, weather, scattering, and other factors. Several methods for satellite image enhancement and restoration have been studied and developed in the literature. The work presented in this thesis, is focused on satellite image restoration by nonlinear statistical filtering techniques. At the first step, we proposed a novel method to restore satellite images using a combination between blind and non-blind restoration techniques. The reason for this combination is to exploit the advantages of each technique used. In the second step, novel statistical image restoration algorithms based on nonlinear filters and the nonparametric multivariate density estimation have been proposed. The nonparametric multivariate density estimation of posterior density is used in the resampling step of the Bayesian bootstrap filter to resolve the problem of loss of diversity among the particles. Finally, we have introduced a new hybrid combination method for image restoration based on the discrete wavelet transform (DWT) and the proposed algorithms in step two, and, we have proved that the performance of the combined method is better than the performance of the DWT approach in the reduction of noise in degraded satellite images.