Approved for public release; distribution in unlimited. / Some battlefield models have a component in them which models the time it takes for an observer to detect a target. Different observers may have different mean detection times due to various factors such as the type of sensor used, environmental conditions, fatigue of the observer, etc. Two parametric models for the distribution of time to target detection are considered which can incorporate these factors. Maximum likelihood estimation procedures for the parameters are described. Results of simulation experiments to study the small sample behavior of the estimators are presented. / http://archive.org/details/modelsfortargetd00baed / Major, Korean Air Force

Survival analysis is applied on mechanical components installed in gas turbines. We use field experience data collected from repair inspection reports. These data are highly censored since the exact time-to-event is unknown. We only know that it lies before or after the repair inspection time. As event we consider irreparability level of the mechanical components. The aim is to estimate survival functions that depend on the different environmental attributes of the sites where the gas turbines operate. Then, the goal is to use this information to obtain optimal time points for preventive maintenance. Optimal times are calculated with respect to the minimization of a cost function which considers expected costs of preventive and corrective maintenance. Another aim is the investigation of the effect of five different failure modes on the component lifetime. The methods used are based on the Weibull distribution, in particular we apply the Bayesian Weibull AFT model and the Bayesian Generalized Weibull model. The latter is preferable for its greater flexibility and better performance. Results reveal that components from gas turbines located in a heavy industrial environment at a higher distance from sea tend to have shorter lifetime. Then, failure mode A seems to be the most harmful for the component lifetime. The model used is capable of predicting customer-specific optimal replacement times based on the effect of environmental attributes. Predictions can be also extended for new components installed at new customer sites.

abstract: This thesis presents a meta-analysis of lead-free solder reliability. The qualitative analyses of the failure modes of lead- free solder under different stress tests including drop test, bend test, thermal test and vibration test are discussed. The main cause of failure of lead- free solder is fatigue crack, and the speed of propagation of the initial crack could differ from different test conditions and different solder materials. A quantitative analysis about the fatigue behavior of SAC lead-free solder under thermal preconditioning process is conducted. This thesis presents a method of making prediction of failure life of solder alloy by building a Weibull regression model. The failure life of solder on circuit board is assumed Weibull distributed. Different materials and test conditions could affect the distribution by changing the shape and scale parameters of Weibull distribution. The method is to model the regression of parameters with different test conditions as predictors based on Bayesian inference concepts. In the process of building regression models, prior distributions are generated according to the previous studies, and Markov Chain Monte Carlo (MCMC) is used under WinBUGS environment. / Dissertation/Thesis / Masters Thesis Industrial Engineering 2014

Selection of signi?cant genes via expression patterns is important in a microarray problem. Owing to small sample size and large number of variables (genes), the selection process can be unstable. This research proposes a hierarchical Bayesian model for gene (variable) selection. We employ latent variables in a regression setting and use a Bayesian mixture prior to perform the variable selection. Due to the binary nature of the data, the posterior distributions of the parameters are not in explicit form, and we need to use a combination of truncated sampling and Markov Chain Monte Carlo (MCMC) based computation techniques to simulate the posterior distributions. The Bayesian model is ?exible enough to identify the signi?cant genes as well as to perform future predictions. The method is applied to cancer classi?cation via cDNA microarrays. In particular, the genes BRCA1 and BRCA2 are associated with a hereditary disposition to breast cancer, and the method is used to identify the set of signi?cant genes to classify BRCA1 and others. Microarray data can also be applied to survival models. We address the issue of how to reduce the dimension in building model by selecting signi?cant genes as well as assessing the estimated survival curves. Additionally, we consider the wellknown Weibull regression and semiparametric proportional hazards (PH) models for survival analysis. With microarray data, we need to consider the case where the number of covariates p exceeds the number of samples n. Speci?cally, for a given vector of response values, which are times to event (death or censored times) and p gene expressions (covariates), we address the issue of how to reduce the dimension by selecting the responsible genes, which are controlling the survival time. This approach enables us to estimate the survival curve when n << p. In our approach, rather than ?xing the number of selected genes, we will assign a prior distribution to this number. The approach creates additional ?exibility by allowing the imposition of constraints, such as bounding the dimension via a prior, which in e?ect works as a penalty. To implement our methodology, we use a Markov Chain Monte Carlo (MCMC) method. We demonstrate the use of the methodology with (a) di?use large B??cell lymphoma (DLBCL) complementary DNA (cDNA) data and (b) Breast Carcinoma data. Lastly, we propose a mixture of Dirichlet process models using discrete wavelet transform for a curve clustering. In order to characterize these time??course gene expresssions, we consider them as trajectory functions of time and gene??speci?c parameters and obtain their wavelet coe?cients by a discrete wavelet transform. We then build cluster curves using a mixture of Dirichlet process priors.

Selection of signi?cant genes via expression patterns is important in a microarray problem. Owing to small sample size and large number of variables (genes), the selection process can be unstable. This research proposes a hierarchical Bayesian model for gene (variable) selection. We employ latent variables in a regression setting and use a Bayesian mixture prior to perform the variable selection. Due to the binary nature of the data, the posterior distributions of the parameters are not in explicit form, and we need to use a combination of truncated sampling and Markov Chain Monte Carlo (MCMC) based computation techniques to simulate the posterior distributions. The Bayesian model is ?exible enough to identify the signi?cant genes as well as to perform future predictions. The method is applied to cancer classi?cation via cDNA microarrays. In particular, the genes BRCA1 and BRCA2 are associated with a hereditary disposition to breast cancer, and the method is used to identify the set of signi?cant genes to classify BRCA1 and others. Microarray data can also be applied to survival models. We address the issue of how to reduce the dimension in building model by selecting signi?cant genes as well as assessing the estimated survival curves. Additionally, we consider the wellknown Weibull regression and semiparametric proportional hazards (PH) models for survival analysis. With microarray data, we need to consider the case where the number of covariates p exceeds the number of samples n. Speci?cally, for a given vector of response values, which are times to event (death or censored times) and p gene expressions (covariates), we address the issue of how to reduce the dimension by selecting the responsible genes, which are controlling the survival time. This approach enables us to estimate the survival curve when n << p. In our approach, rather than ?xing the number of selected genes, we will assign a prior distribution to this number. The approach creates additional ?exibility by allowing the imposition of constraints, such as bounding the dimension via a prior, which in e?ect works as a penalty. To implement our methodology, we use a Markov Chain Monte Carlo (MCMC) method. We demonstrate the use of the methodology with (a) di?use large B??cell lymphoma (DLBCL) complementary DNA (cDNA) data and (b) Breast Carcinoma data. Lastly, we propose a mixture of Dirichlet process models using discrete wavelet transform for a curve clustering. In order to characterize these time??course gene expresssions, we consider them as trajectory functions of time and gene??speci?c parameters and obtain their wavelet coe?cients by a discrete wavelet transform. We then build cluster curves using a mixture of Dirichlet process priors.

A análise de confiabilidade desempenha um papel fundamental para estudos de durabilidade e otimização de tempos de reparo em sistemas reparáveis. Equipamentos como colhedoras de cana-de-açúcar que após a falha e um reparo voltam a exercer sua função objetivo são classificados como sistemas reparáveis. O objetivo deste trabalho consistiu em propor alternativas de modelagem para sistemas complexos, que apresentam grande variabilidade no comportamento da função intensidade de falha. Foi proposta a nova distribuição odd log-logística Weibull flexível generalizada (GOLLFW) e um modelo de regressão Weibull aplicado ao processo lei de potência usado para analisar sistemas reparáveis. Para a nova distribuição foi apresentada a família de distribuições odd log-logística generalizada, realizado um estudo de simulação para verificar algumas propriedades dos estimadores de máxima verossimilhança e incluídas covariáveis na análise dos tempos de falha através do modelo de regressão GOLLFW. Para a análise de regressão considerando os sistemas reparáveis, foram apresentados os principais modelos de contagem para um único sistema reparável e realizado a análise deles de forma separada e, em seguida, foram considerados mais de dois sistemas e acrescentado um modelo de regressão Weibull ao processo lei de potência (PLP). A característica de bimodalidade da distribuição GOLLFW garantiu a adequabilidade e um melhor ajuste aos dados. Já a inclusão de covariáveis através do modelo de regressão Weibull no PLP permitiu modelar sistemas que antes somente os processos de contagens tradicionais, processo lei de potência e processo de renovação, não se adequariam bem. / The confiability analysis carries out an important role for durability studies and optimization of repair time in repairable systems. Repairable systems are equipments that returns to execute its function after a fail, for example, sugarcane harvester. This work aimed to propose modeling alternatives for complex systems with great variability in the behaviour of fail intensity function. It was proposed a new distribution on generalized odd log-logistic flexible Weibull (GOLLFW) and an Weibull regression model applied to potential law used to analyze repairable systems.It was presented the distribution family generalized odd log-logistic, was carried out a simulation study to verify some properties of maximum likelihood estimators and was included covariables in the fail time by regression model GOLLFW. To the regression analysis considering repairable systems, it was presented the main counting models for a single repairable system and it was performed an analysis of each model singly, then, it was considered more than two systems and it was added a Weibull regression model to the potential law process (PLP). The bimodality characteristic of GOLLFW distribution guaranteed the suitability and a better adjust to tested datas. While, the inclusion of covariables by regression model GOLLFW in the PLP allowed to model systems which traditionals counting process, PLP and renewal process, would not fit well.

Em aplicações na área de análise de sobrevivência, é freqüente a ocorrência de função de taxa de falha em forma de U ou unimodal, isto e, funções não-monótonas. Os modelos de regressão comumente usados para dados de sobrevivência são log-Weibull, função de taxa de falha monótona, e log-logística, função de taxa de falha decrescente ou unimodal. Um dos objetivos deste trabalho e propor os modelos de regressão, em forma de locação e escala, log-Weibull estendida que apresenta função de taxa de falha em forma de U e log- Burr XII que tem como caso particular o modelo de regressão log-logística. Considerando dados censurados, foram utilizados três métodos para estimação dos parâmetros, a saber, máxima verossimilhança, bayesiana e jackkinife. Para esses modelos foram calculadas algumas medidas de diagnósticos de influência local e global. Adicionalmente, desenvolveu-se uma análise de resíduos baseada no resíduo tipo martingale. Para diferentes parâmetros taxados, tamanhos de amostra e porcentagens de censuras, várias simulações foram feitas para avaliar a distribuição empírica do resíduo tipo martingale e compará-la com a distribuição normal padrão. Esses estudos sugerem que a distribuição empírica do resíduo tipo martingale para o modelo de regressão log-Weibull estendida com dados censurados aproxima-se de uma distribuição normal padrão quando comparados com outros resíduos considerados neste estudo. Para o modelo de regressão log-Burr XII, foi proposta uma modificação no resíduo tipo martingale baseada no estudo de simulação para obter concordância com a distribuição normal padrão. Conjuntos de dados reais foram utilizados para ilustrar a metodologia desenvolvida. Também pode ocorrer que em algumas aplicações a suposição de independência dos tempos de sobrevivência não é válida. Assim, outro objetivo deste trabalho é introduzir um modelo de regressão log-Burr XII com efeito aleatório para o qual foi proposto um método de estimação para os parâmetros baseado no algoritmo EM por Monte Carlo. Por fim, foi desenvolvido um novo modelo probabilístico denominado de beta Weibull modificado que apresenta cinco parâmetros. A vantagem desse novo modelo é a flexibilidade em acomodar várias formas da função de taxa de falha, por exemplo, U e unimodal, e mostrou-se útil na discriminação entre alguns modelos probabilísticos alternativos. O método de máxima verossimilhança e proposto para estimar os parâmetros desta distribuição. A matriz de informação observada foi calculada. Um conjunto de dados reais é usado para ilustrar a aplicação da nova distribuição / In survival analysis applications, the failure rate function may have frequently unimodal or bathtub shape, that is, non-monotone functions. The regression models commonly used for survival studies are log-Weibull, monotone failure rate function shape, and log-logistic, decreased or unimodal failure rate function shape. In the first part of this thesis, we propose location-scale regression models based on an extended Weibull distribution for modeling data with bathtub-shaped failure rate function and on a Burr XII distribution as an alternative to the log-logistic regression model. Assuming censored data, we consider a classical analysis, a Bayesian analysis and a jackknife estimator for the parameters of the proposed models. For these models, we derived the appropriate matrices for assessing the local influence on the parameter estimates under diferent perturbation schemes, and we also presented some ways to perform global influence. Additionally, we developed residual analy- sis based on the martingale-type residual. For di®erent parameter settings, sample sizes and censoring percentages, various simulation studies were performed and the empirical distribution of the martingale-type residual was displayed and compared with the standard normal distribution. These studies suggest that the empirical distribution of the martingale-type residual for the log-extended Weibull regression model with data censured present a high agreement with the standard normal distribution when compared with other residuals considered in these studies. For the log-Burr XII regression model, it was proposed a change in the martingale-type residual based on some studies of simulation in order to obtain an agreement with the standard normal distribution. Some applications to real data illustrate the usefulness of the methodology developed. It can also happen in some applications that the assumption of independence of the times of survival is not valid, so it was added to the log-Burr XII regression model of random exects for which an estimate method was proposed for the parameters based on the EM algorithm for Monte Carlo simulation. Finally, a five- parameter distribution so called the beta modified Weibull distribution is defined and studied. The advantage of that new distribution is its flexibility in accommodating several forms of the failure rate function, for instance, bathtub-shaped and unimodal shape, and it is also suitable for testing goodness-of-fit of some special sub-models. The method of maximum likelihood is used for estimating the model parameters. We calculate the observed information matrix. A real data set is used to illustrate the application of the new distribution.

A análise de confiabilidade desempenha um papel fundamental para estudos de durabilidade e otimização de tempos de reparo em sistemas reparáveis. Equipamentos como colhedoras de cana-de-açúcar que após a falha e um reparo voltam a exercer sua função objetivo são classificados como sistemas reparáveis. O objetivo deste trabalho consistiu em propor alternativas de modelagem para sistemas complexos, que apresentam grande variabilidade no comportamento da função intensidade de falha. Foi proposta a nova distribuição odd log-logística Weibull flexível generalizada (GOLLFW) e um modelo de regressão Weibull aplicado ao processo lei de potência usado para analisar sistemas reparáveis. Para a nova distribuição foi apresentada a família de distribuições odd log-logística generalizada, realizado um estudo de simulação para verificar algumas propriedades dos estimadores de máxima verossimilhança e incluídas covariáveis na análise dos tempos de falha através do modelo de regressão GOLLFW. Para a análise de regressão considerando os sistemas reparáveis, foram apresentados os principais modelos de contagem para um único sistema reparável e realizado a análise deles de forma separada e, em seguida, foram considerados mais de dois sistemas e acrescentado um modelo de regressão Weibull ao processo lei de potência (PLP). A característica de bimodalidade da distribuição GOLLFW garantiu a adequabilidade e um melhor ajuste aos dados. Já a inclusão de covariáveis através do modelo de regressão Weibull no PLP permitiu modelar sistemas que antes somente os processos de contagens tradicionais, processo lei de potência e processo de renovação, não se adequariam bem. / The confiability analysis carries out an important role for durability studies and optimization of repair time in repairable systems. Repairable systems are equipments that returns to execute its function after a fail, for example, sugarcane harvester. This work aimed to propose modeling alternatives for complex systems with great variability in the behaviour of fail intensity function. It was proposed a new distribution on generalized odd log-logistic flexible Weibull (GOLLFW) and an Weibull regression model applied to potential law used to analyze repairable systems.It was presented the distribution family generalized odd log-logistic, was carried out a simulation study to verify some properties of maximum likelihood estimators and was included covariables in the fail time by regression model GOLLFW. To the regression analysis considering repairable systems, it was presented the main counting models for a single repairable system and it was performed an analysis of each model singly, then, it was considered more than two systems and it was added a Weibull regression model to the potential law process (PLP). The bimodality characteristic of GOLLFW distribution guaranteed the suitability and a better adjust to tested datas. While, the inclusion of covariables by regression model GOLLFW in the PLP allowed to model systems which traditionals counting process, PLP and renewal process, would not fit well.

摘要 本文主要考慮群集樣本(clustered samples)的存活分析，而每一群集中又分為兩種組別(groups)。假定同群集同組別內的個體共享相同但不可觀測的隨機脆弱性(frailty)，因此面臨的是雙變量脆弱性變數的多變量存活資料。首先，驗證雙變量脆弱性對雙變量對數存活時間及雙變量存活時間之相關係數所造成的影響。接著，假定雙變量脆弱性服從雙變量對數常態分配，條件存活時間模式為韋伯迴歸模式，我們利用EM法則，推導出雙變量脆弱性之多變量存活模式中母數的估計方法。 關鍵詞：雙變量脆弱性，Weibull迴歸模式，對數常態分配，EM法則 / Abstract Consider survival analysis for clustered samples, where each cluster contains two groups. Assume that individuals within the same cluster and the same group share a common but unobservable random frailty. Hence, the focus of this work is on bivariate frailty model in analysis of multivariate survival data. First, we derive expressions for the correlation between the two survival times to show how the bivariate frailty affects these correlation coefficients. Then, the bivariate log-normal distribution is used to model the bivariate frailty. We modified EM algorithm to estimate the parameters for the Weibull regression model with bivariate log-normal frailty. Key words：bivariate frailty, Weibull regression model, log-normal distribution, EM algorithm.

Submitted by (ana.araujo@ufrpe.br) on 2016-05-20T14:30:00Z No. of bitstreams: 1 Andrea Andrade Prudente.pdf: 1364424 bytes, checksum: 52db48248a4f42fd96b6ee53463083eb (MD5) / Made available in DSpace on 2016-05-20T14:30:00Z (GMT). No. of bitstreams: 1 Andrea Andrade Prudente.pdf: 1364424 bytes, checksum: 52db48248a4f42fd96b6ee53463083eb (MD5) Previous issue date: 2009-03-18 / Conselho Nacional de Pesquisa e Desenvolvimento Científico e Tecnológico - CNPq / The main objective in this dissertation is to derive expressions for the second-order biases of the maximum likelihood estimators of the parameters of the Weibull generalized linear model (WGLM), which are useful to define corrected estimators. In order to reduce the bias of these estimators in finite sample sizes, the method of bias correction introduced by Cox and Snell (1968) was used. The new model adopts a link function which relates the vector of scale parameters of the Weibull distribution to a linear predictor. As a second objective, a revision of the normal non-linear models was also presented, including the method of least squares for estimating the parameters, some asymptotic results, measures of nonlinearity and diagnostic techniques, because in contrast to linear models, quality and, especially, the validity of their fits are evaluated not only by means of regression diagnostics, but also with the extent of the non-linear behavior. Finally, a brief description of generalized linear models (GLM) is given and the applicability of the model range. Real data sets were analyzed to demonstrate the applicability of the proposed models. These tests were conducted in the R environment for programming, data analysis, andgraphics. / Esta dissertação tem como objetivo principal apresentar expressões para os vieses de segunda ordem dos estimadores de máxima verossimilhança dos parâmetros do modelo linear generalizado de Weibull (MLGW), utilizando-as para obter estimadores corrigidos. Com o intuito de reduzir os vieses destes estimadores, em amostras de tamanho finito, utilizou-se a correção do viés pelo uso da equação de Cox e Snell (1968). Esse modelo permite a utilização de uma função de ligação para relacionar o vetor dos parâmetros de escala da distribuição de Weibull (parte da média) ao preditor linear. Um objetivo secundário foi revisar os modelos normais não-lineares, contemplando o método de mínimos quadrados para estimação dos seus parâmetros, alguns resultados assintóticos, medidas de não-linearidade e técnicas de diagnóstico, pois ao contrário dos modelos lineares, a qualidade e, principalmente, a validade dos seus ajustes são avaliadas não só por meio de diagnósticos de regressão, mas pela extensão do comportamento nãolinear. Por fim, foi apresentada, também, uma sucinta descrição dos modelos lineares generalizados (MLG) e a aplicabilidade do modelo gama. Dados reais foram analisados para demonstrar a aplicabilidade dos modelos propostos. Estas análises foram realizadas no ambiente de programação, análise de dados e gráficos R.