Submitted by Luciana Ferreira (lucgeral@gmail.com) on 2017-05-02T15:17:50Z No. of bitstreams: 2 Dissertação - Mário Fernando de Sousa - 2016.pdf: 645506 bytes, checksum: d6fd190570fce6feeb390cfeaf50032f (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) / Approved for entry into archive by Luciana Ferreira (lucgeral@gmail.com) on 2017-05-02T15:18:06Z (GMT) No. of bitstreams: 2 Dissertação - Mário Fernando de Sousa - 2016.pdf: 645506 bytes, checksum: d6fd190570fce6feeb390cfeaf50032f (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) / Made available in DSpace on 2017-05-02T15:18:06Z (GMT). No. of bitstreams: 2 Dissertação - Mário Fernando de Sousa - 2016.pdf: 645506 bytes, checksum: d6fd190570fce6feeb390cfeaf50032f (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) Previous issue date: 2016-12-06 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - CAPES / This work aims to fill a gap in the literature on modeling asymmetric and censored data. The main objective is to provide a contribution by developing two models, which will be presented in two papers, respectively. In the first paper, we develop the tobit-Birnbaum-Saunders model, a variation of the standard tobit model. We discuss estimation based on the maximum likelihood method, residuals, diagnostic techniques and an empirical application. In the second paper, we propose the use of a mixture between the Birnbaum-Saunders and Bernoulli distributions. The objective is to generalize the tobit-Birnbaum-Saunders model in order to consider the possibility of partial observations below a cutoff point. For the mixture model, we carry out a Monte Carlo simulation study and an empirical application. The results show that, in both cases, the Birnbaum-Saunders distribution provides the best results. / Este trabalho visa preencher uma lacuna existente na literatura pertinente à modelagem de dados assimétricos e censurados. O objetivo principal é oferecer uma contribuição via o desenvolvimento de dois modelos, os quais serão apresentados em dois artigos. No primeiro artigo é proposto o modelo tobit-Birnbaum-Saunders, ou seja, uma variação do modelo tobit clássico, com estimação baseada no método de máxima verossimilhança, resíduos, técnicas de diagnóstico e uma aplicação a dados reais. No segundo artigo é abordada a utilização de um modelo de mistura entre as distribuições Birnbaum-Saunders e Bernoulli, de modo a generalizar o modelo tobit-Birnbaum-Saunders e considerar a possibilidade de observações parciais abaixo do ponto de corte. Para o modelo de mistura são realizadas simulações de Monte Carlo e uma aplicação a dados reais. Os resultados mostram que, em ambos os casos, a distribuição Birnbaum-Saunders oferece os melhores resultados.

No presente trabalho, e proposto um modelo de regressão utilizando a distribuição gama generalizada exponenciada (GGE) para dados censurados, esta nova distribuição e uma extensão da distribuição gama generalizada. A distribuição GGE (CORDEIRO et al., 2009) que tem quatro parâmetros pode modelar dados de sobrevivência quando a função de risco tem forma crescente, decrescente, forma de U e unimodal. Neste trabalho apresenta-se uma expansão natural da distribuição GGE para dados censurados, esta distribuição desperta o interesse pelo fato de representar uma família paramétrica que possui como casos particulares outras distribuições amplamente utilizadas na analise de dados de tempo de vida, como as distribuições gama generalizada (STACY, 1962), Weibull, Weibull exponenciada (MUDHOLKAR et al., 1995, 1996), exponencial exponenciada (GUPTA; KUNDU, 1999, 2001), Rayleigh generalizada (KUNDU; RAKAB, 2005), dentre outras, e mostra-se útil na discriminação entre alguns modelos probabilísticos alternativos. Considerando dados censurados, e abordado o método de máxima verossimilhança para estimar os parâmetros do modelo proposto. Outra proposta deste trabalho e introduzir um modelo de regressão log-gama generalizado exponenciado com efeito aleatório. Por fim, são apresentadas três aplicações para ilustrar a distribuição proposta. / In the present study, we propose a regression model using the exponentiated generalized gama (EGG) distribution for censored data, this new distribution is an extension of the generalized gama distribution. The EGG distribution (CORDEIRO et al., 2009) that has four parameters it can model survival data when the risk function is increasing, decreasing, form of U and unimodal-shaped. In this work comes to a natural expansion of the EGG distribution for censored data, is awake distribution the interest for the fact of representing a parametric family that has, as particular cases, other distributions which are broadly used in lifetime data analysis, as the generalized gama (STACY, 1962), Weibull, exponentiated Weibull (MUDHOLKAR et al., 1995, 1996), exponentiated exponential (GUPTA; KUNDU, 1999, 2001), generalized Rayleigh (KUNDU; RAKAB, 2005), among others, and it is shown useful in the discrimination among some models alternative probabilistics. Considering censored data, the maximum likelihood estimator is considered for the proposed model parameters. Another proposal of this work was to introduce a log-exponentiated generalized gamma regression model with random eect. Finally, three applications were presented to illustrate the proposed distribution.

Esta tese visa propor extensões dos modelos de regressão quantílica bayesianos, considerando dados de proporção com inflação de zeros, e também dados censurados no zero. Inicialmente, é sugerida uma análise de observações influentes, a partir da representação por mistura localização-escala da distribuição Laplace assimétrica, em que as distribuições a posteriori das variáveis latentes são comparadas com o intuito de identificar possíveis observações aberrantes. Em seguida, é proposto um modelo de duas partes para analisar dados de proporção com inflação de zeros ou uns, estudando os quantis condicionais e a probabilidade da variável resposta ser igual a zero. Além disso, são propostos modelos de regressão quantílica bayesiana para dados contínuos com um componente discreto no zero, em que parte dessas observações é suposta censurada. Esses modelos podem ser considerados mais completos na análise desse tipo de dados, uma vez que a probabilidade de censura é verificada para cada quantil de interesse. E por último, é considerada uma aplicação desses modelos com correlação espacial, para estudar os dados da eleição presidencial no Brasil em 2014. Nesse caso, os modelos de regressão quantílica são capazes de incorporar essa informação espacial a partir do processo Laplace assimétrico. Para todos os modelos propostos foi desenvolvido um pacote do software R, que está exemplificado no apêndice. / This thesis aims to propose extensions of Bayesian quantile regression models, considering proportion data with zero inflation, and also censored data at zero. Initially, it is suggested an analysis of influential observations, based on the location-scale mixture representation of the asymmetric Laplace distribution, where the posterior distribution of the latent variables are compared with the goal of identifying possible outlying observations. Next, a two-part model is proposed to analyze proportion data with zero or one inflation, studying the conditional quantile and the probability of the response variable being equal to zero. Following, Bayesian quantile regression models are proposed for continuous data with a discrete component at zero, where part of these observations are assumed censored. These models may be considered more complete in the analysis of this type of data, as the censoring probability varies with the quantiles of interest. For last, it is considered an application of these models with spacial correlation, in order to study the data about the last presidential election in Brazil in 2014. In this example, the quantile regression models are able to incorporate spatial dependence with the asymmetric Laplace process. For all the proposed models it was developed a R package, which is exemplified in the appendix.

Cette thèse présente divers problèmes d'estimation fonctionnelle adaptative par sélection d'estimateurs en projection ou à noyaux, utilisant des critères inspirés à la fois de la sélection de modèles et des méthodes de Lepski. Le point commun de nos travaux est l'utilisation de données transformées et/ou incomplètes. La première partie est consacrée à une procédure d'estimation par "déformation'', dont la pertinence est illustrée pour l'estimation des fonctions suivantes : régression additive et multiplicative, densité conditionnelle, fonction de répartition dans un modèle de censure par intervalle, risque instantané pour des données censurées à droite. Le but est de reconstruire une fonction à partir d'un échantillon de couples aléatoires (X,Y). Nous utilisons les données déformées (ф(X),Y) pour proposer des estimateurs adaptatifs, où ф est une fonction bijective que nous estimons également (par exemple la fonction de répartition de X). L'intérêt est double : d'un point de vue théorique, les estimateurs ont des propriétés d'optimalité au sens de l'oracle ; d'un point de vue pratique, ils sont explicites et numériquement stables. La seconde partie s'intéresse à un problème à deux échantillons : nous comparons les distributions de deux variables X et Xₒ au travers de la densité relative, définie comme la densité de la variable Fₒ(X) (Fₒ étant la répartition de Xₒ). Nous construisons des estimateurs adaptatifs, à partir d'un double échantillon de données, possiblement censurées. Des bornes de risque non-asymptotiques sont démontrées, et des vitesses de convergences déduites. / This thesis presents various problems of adaptive functional estimation, using projection and kernel methods, and criterions inspired both by model selection and Lepski's methods. The common point of the studied statistical setting is to deal with transformed and/or incomplete data. The first part proposes a method of estimation with a "warping" device which permits to handle the estimation of functions such as additive and multiplicative regression, conditional density, hazard rate based on randomly right-censored data, and cumulative distribution function from current-status data. The aim is to estimate a function from a sample of random variable (X,Y). We use the warped data (ф(X),Y), to propose adaptive estimators, where ф is a one-to-one function that we also estimate (e.g. the cumulative distribution function of X). The interest is twofold. From the theoretical point of view, the estimators are optimal in the oracle sense. From the practical point of view, they can be easily computed, thanks to their simple explicit expression. The second part deals with a two-sample problem : we compare the distribution of two variables X and Xₒ by studying the relative density, defined as the density of Fₒ(X) (Fₒ is the c.d.f. of Xₒ). We build adaptive estimators, from a double data-sample, possibly censored. Non-asymptotic risk bounds are proved, and convergence rates are also derived.

Dans cette thèse, nous proposons une démarche méthodologique permettant de simuler le besoin en équipement de rechange pour une flotte de véhicules. Les systèmes se dégradent avec l’âge ou l’usage, et sont défaillants lorsqu’ils ne remplissent plus leur mission. L’usager a alors besoin d’une assurance que le système soit opérationnel pendant sa durée de vie utile. Un contrat de soutien oblige ainsi l’industriel à remédier à une défaillance et à maintenir le système en condition opérationnelle durant la durée du contrat. Ces dernières années, la mondialisation et l’évolution rapide des technologies obligent les constructeurs à proposer des offres de contrat de maintenance bien au-delà de la vie utile des équipements. La gestion de contrat de soutien ou d’extension de soutien requiert la connaissance de la durée de vie des équipements, mais aussi des conditions d’usages des véhicules, dépendant du client. L’analyse des retours clientèle ou des RetEx est alors un outil important d’aide à la décision pour l’industriel. Cependant ces données ne sont pas homogènes et sont très fortement censurées, ce qui rend les estimations difficiles. La plupart du temps, cette variabilité n’est pas observée mais doit cependant être prise en compte sous peine d’erreur de décision. Nous proposons dans cette thèse de modéliser l’hétérogénéité des durées de vie par un modèle de mélange et de concurrence de deux lois de Weibull. On propose une méthode d’estimation des paramètres capable d’être performante malgré la forte présence de données censurées.Puis, nous faisons appel à une méthode de classification non supervisée afin d’identifier des profils d’utilisation des véhicules. Cela nous permet alors de simuler les besoins en pièces de rechange pour une flotte de véhicules pour la durée du contrat ou pour une extension de contrat. / This thesis gathers methodologicals contributions to simulate the need of replacement equipment for a vehile fleet. Systems degrade with age or use, and fail when they do not fulfill their mission. The user needs an assurance that the system is operational during its useful life. A support contract obliges the manufacturer to remedy a failure and to keep the system in operational condition for the duration of the MCO contract.The management of support contracts or the extension of support requires knowledge of the equipment lifetime and also the uses condition of vehicles, which depends on the customer. The analysis of customer returns or RetEx is then an important tool to help support the decision of the industrial. In reliability or warranty analysis, engineers must often deal with lifetimes data that are non-homogeneous. Most of the time, this variability is unobserved but has to be taken into account for reliability or warranty cost analysis.A further problem is that in reliability analysis, the data is heavily censored which makes estimations more difficult. We propose to consider the heterogeneity of lifetimes by a mixture and competition model of two Weibull laws. Unfortunately, the performance of classical estimation methods (maximum of likelihood via EM, Bayes approach via MCMC) is jeopardized due to the high number of parameters and the heavy censoring.To overcome the problem of heavy censoring for Weibull mixture parameters estimation, we propose a Bayesian bootstrap method, called Bayesian RestorationMaximization.We use an unsupervised clustering method to identify the profiles of vehicle uses. Our method allows to simulate the needs of spare parts for a vehicles fleet for the duration of the contract or for a contract extension.

Neste trabalho, foi proposta uma nova família de distribuições, a qual permite modelar dados de sobrevivência quando a função de risco tem formas unimodal e U (banheira). Ainda, foram consideradas as modificações das distribuições Weibull, Fréchet, half-normal generalizada, log-logística e lognormal. Tomando dados não-censurados e censurados, considerou-se os estimadores de máxima verossimilhança para o modelo proposto, a fim de verificar a flexibilidade da nova família. Além disso, um modelo de regressão locação-escala foi utilizado para verificar a influência de covariáveis nos tempos de sobrevida. Adicionalmente, conduziu-se uma análise de resíduos baseada nos resíduos deviance modificada. Estudos de simulação, utilizando-se de diferentes atribuições dos parâmetros, porcentagens de censura e tamanhos amostrais, foram conduzidos com o objetivo de verificar a distribuição empírica dos resíduos tipo martingale e deviance modificada. Para detectar observações influentes, foram utilizadas medidas de influência local, que são medidas de diagnóstico baseadas em pequenas perturbações nos dados ou no modelo proposto. Podem ocorrer situações em que a suposição de independência entre os tempos de falha e censura não seja válida. Assim, outro objetivo desse trabalho é considerar o mecanismo de censura informativa, baseado na verossimilhança marginal, considerando a distribuição log-odd log-logística Weibull na modelagem. Por fim, as metodologias descritas são aplicadas a conjuntos de dados reais. / In this study, a new family of distributions was proposed, which allows to model survival data when the function of risk has unimodal shapes and U (bathtub). Modifications of the Weibull, Fréchet, generalized half-normal, log-logistic and lognormal distributions were considered. Taking censored and non-censored data, we consider the maximum likelihood estimators for the proposed model, in order to check the flexibility of the new family. Also, it was considered a location-scale regression model, to verify the influence of covariates on survival times. Additionally, a residual analysis was conducted based on modified deviance residuals. For different parameters fixed, percentages of censoring and sample sizes, several simulation studies were performed with the objective of verify the empirical distribution of the martingale type and modified deviance residuals. To detect influential observations, measures of local influence were used, which are diagnostic measures based on small perturbations in the data or in the proposed model. It can occur situations in which the assumption of independence between the failure and censoring times is not valid. Thus, another objective of this work is to consider the informative censoring mechanism based on the marginal likelihood, considering the log-odd log-logistic Weibull distribution in modelling. Finally, the methodologies described are applied to sets of real data.

No description available.

本文所探討的中心為貝氏模型運用於加速壽命試驗，並且假設受測項目之壽命服從Weibull分配。加速實驗環境有三種，其中第二種環境代表正常狀態，採用加速壽命試驗的方式涵蓋了三種：固定應力、漸進之逐步應力和變量曲線之逐步應力。對於先驗參數，並不是直接給予特定的值，而是透過專家評估，給定各種環境之下的產品可靠度之中位數或百分位數，再利用這些資訊經過數值運算解出先驗參數。資料的型態分成兩種，一為區間資料，另一為型一設限資料，透過蒙地卡羅法模擬出後驗分配，並且估計正常環境狀態的可靠度。 / This article develops a Bayes inference model for accelerated life testing assuming failure times at each stress level are Weibull distributed. Using the approach, there are three stressed to be used, and the three testing scenarios to be adapted are as follows：fixed-stress, progressive step-stress and profile step-stress. Prior information is used to indirectly define a multivariate prior distribution for the scale parameters at the various stress levels. The inference procedure accommodates both the interval data sampling strategy and type I censored sampling strategy for the collection of ALT test data. The inference procedure uses the well-known Markov Chain Monte Carlo methods to derive posterior approximations.

Neste trabalho, foi proposta uma nova família de distribuições, a qual permite modelar dados de sobrevivência quando a função de risco tem formas unimodal e U (banheira). Ainda, foram consideradas as modificações das distribuições Weibull, Fréchet, half-normal generalizada, log-logística e lognormal. Tomando dados não-censurados e censurados, considerou-se os estimadores de máxima verossimilhança para o modelo proposto, a fim de verificar a flexibilidade da nova família. Além disso, um modelo de regressão locação-escala foi utilizado para verificar a influência de covariáveis nos tempos de sobrevida. Adicionalmente, conduziu-se uma análise de resíduos baseada nos resíduos deviance modificada. Estudos de simulação, utilizando-se de diferentes atribuições dos parâmetros, porcentagens de censura e tamanhos amostrais, foram conduzidos com o objetivo de verificar a distribuição empírica dos resíduos tipo martingale e deviance modificada. Para detectar observações influentes, foram utilizadas medidas de influência local, que são medidas de diagnóstico baseadas em pequenas perturbações nos dados ou no modelo proposto. Podem ocorrer situações em que a suposição de independência entre os tempos de falha e censura não seja válida. Assim, outro objetivo desse trabalho é considerar o mecanismo de censura informativa, baseado na verossimilhança marginal, considerando a distribuição log-odd log-logística Weibull na modelagem. Por fim, as metodologias descritas são aplicadas a conjuntos de dados reais. / In this study, a new family of distributions was proposed, which allows to model survival data when the function of risk has unimodal shapes and U (bathtub). Modifications of the Weibull, Fréchet, generalized half-normal, log-logistic and lognormal distributions were considered. Taking censored and non-censored data, we consider the maximum likelihood estimators for the proposed model, in order to check the flexibility of the new family. Also, it was considered a location-scale regression model, to verify the influence of covariates on survival times. Additionally, a residual analysis was conducted based on modified deviance residuals. For different parameters fixed, percentages of censoring and sample sizes, several simulation studies were performed with the objective of verify the empirical distribution of the martingale type and modified deviance residuals. To detect influential observations, measures of local influence were used, which are diagnostic measures based on small perturbations in the data or in the proposed model. It can occur situations in which the assumption of independence between the failure and censoring times is not valid. Thus, another objective of this work is to consider the informative censoring mechanism based on the marginal likelihood, considering the log-odd log-logistic Weibull distribution in modelling. Finally, the methodologies described are applied to sets of real data.

A device that is accompanied by an irreversible chemical reaction or physical destruction and could no longer function properly after performing its intended function is referred to as a one-shot device. One-shot device test data differ from typical data obtained by measuring lifetimes in standard life-tests. Due to the very nature of one-shot devices, actual lifetimes of one-shot devices under test cannot be observed, and they are either left- or right-censored. In addition, a one-shot device often has multiple components that could cause the failure of the device. The components are coupled together in the manufacturing process or assembly, resulting in the failure modes possessing latent heterogeneity and dependence. Frailty models enable us to describe the influence of common, but unobservable covariates, on the hazard function as a random effect in a model and also provide an easily understandable interpretation. In this thesis, we develop some inferential results for one-shot device testing data with gamma frailty model. We first develop an efficient expectation-maximization (EM) algorithm for determining the maximum likelihood estimates of model parameters of a gamma frailty model with exponential lifetime distributions for components based on one-shot device test data with multiple failure modes, wherein the data are obtained from a constant-stress accelerated life-test. The maximum likelihood estimate of the mean lifetime of $k$-out-of-$M$ structured one-shot devices under normal operating conditions is also presented. In addition, the asymptotic variance–covariance matrix of the maximum likelihood estimates is derived, which is then used to construct asymptotic confidence intervals for the model parameters. The performance of the proposed inferential methods is finally evaluated through Monte Carlo simulations and then illustrated with a numerical example. A gamma frailty model with Weibull baseline hazards is considered next for fitting one-shot device testing data. The Weibull baseline hazards enable us to analyze time-varying failure rates more accurately, allowing for a deeper understanding of the dynamic nature of system's reliability. We develop an EM algorithm for estimating the model parameters utilizing the complete likelihood function. A detailed simulation study evaluates the performance of the Weibull baseline hazard model with that of the exponential baseline hazard model. The introduction of shape parameters in the component's lifetime distribution within the Weibull baseline hazard model offers enhanced flexibility in model fitting. Finally, Bayesian inference is then developed for the gamma frailty model with exponential baseline hazard for one-shot device testing data. We introduce the Bayesian estimation procedure using Markov chain Monte Carlo (MCMC) technique for estimating the model parameters as well as for developing credible intervals for those parameters. The performance of the proposed method is evaluated in a simulation study. Model comparison between independence model and the frailty model is made using Bayesian model selection criterion. / Thesis / Candidate in Philosophy