[pt] O uso da técnica bootstrap para construção de intervalos de confiança e testes de hipóteses vem aumentando consideravelmente desde seu surgimento, em 1979, devido principalmente ao rápido avanço computacional ocorrido nas últimas décadas. Neste trabalho utilizamos o bootstrap paramétrico e não-paramétrico para estudar o comportamento dos hiperparametros em modelos de espaço de estados nos casos de nível e tendência linear locais. Intervalos de confiança baseados em quatro métodos bootstrap diferentes são calculados e comparados quanto à probabilidade de cobertura, produzindo resultados satisfatórios. Constatamos também a eficiência dos testes boopstrap para os casos em que os hiperparâmetros caem no limite do espaço paramétrico, situação que inviabiliza o uso dos testes clássicos por violar uma das condições de regularidade do estimador de máxima verossimilhança. / [en] Bootstrap procedures to calculate confidence intervals and hypotheses tests had considerable growth since its first appearance, in 1979, mostly due to the rapid computational developments that occurred in the last decades. In this work we employ the parametric and nonparametric boorstrap to study the behaviour of hyperparameters in state-space models in the case of local level and linear trend. Confidence intervals based on four different bootstrap methods are computed and compared using the coverage probabilities, with satisfactory results. We also verify the efficiency of bootstrap tests in cases where the hyperparameters lie on the boundary of the parameter space, situation that makes the classical tests inadequate to use, as it violates one of the regularity conditions of the maximum likelihood estimator.

Orientador: Rubén Augusto Romero Lázaro / Resumo: Tradicionalmente, o problema de Planejamento da Expansão de Sistemas de Transmissão (PEST) é solucionado considerando apenas um único cenário de geração, embora sistemas elétricos reais operem em diferentes cenários de geração. Nessa pesquisa são propostos modelos matemáticos para resolver o problema de PEST, considerando múltiplos cenários de geração de forma que o plano de expansão obtido permita uma operação adequada do sistema. No modelo proposto, o custo de investimento é maior em relação aos planos de expansão encontrados pelo planejamento tradicional, que considera apenas um cenário de geração. Para reduzir o correspondente custo de investimento são apresentadas estratégias eficientes para encontrar planos de expansão para o problema de PEST considerando múltiplos cenários. As estratégias utilizadas foram: permitir pequenos cortes de carga; permitir o deslocamento do nível de geração em uma pequena faixa de geração mínima e máxima em relação à geração ideal e permitir pequenas sobrecargas nas linhas de transmissão. Adicionalmente, uma combinação entre essas estratégias é apresentada e o problema PEST também foi resolvido para o planejamento multiestágio, considerando múltiplos cenários de geração. O método proposto foi implementado com o uso da linguagem de modelagem algébrica AMPL e resolvido com o uso do solver comercial CPLEX. Os resultados encontrados correspondem à propostas de solução que são válidas para diferentes cenários de geração e apresentam diferentes alt... (Resumo completo, clicar acesso eletrônico abaixo) / Abstract: Traditionally, the Transmission Network Expansion Problem is solved considering only a single generation scenario. However, a real power system operates in different generation scenarios. This work presents the disjunctive linear model for the Transmission Network Expansion Problem considering multiple generation scenarios to provide a single expansion plan, which must operate in a appropriate way in each one of the different scenarios. The investment cost of the proposed model is greater in relation to the traditional expansion plans, that consider single generation scenario. In order to reduce the investment costs, efficient strategies are presented to find the expansion plans for multiples scenarios. Therefore those strategies are: allow small load cuts; allow generation level displacement in a narrow generation range in relation to the ideal one; and allow small overload in the transmission lines. Moreover, a combination between those strategies is shown and the Transmission Network Expansion Problem was also solved for multistage planning for multiple generation scenarios. The proposed method was implemented using A Mathematical Programming Language (AMPL) and the commercial solver CPLEX. The results were of optimal quality, considering the characteristics of the used solver, and they were compared with methods found in the specialized literature. / Doutor

abstract: The tools developed for the use of investigating dynamical systems have provided critical understanding to a wide range of physical phenomena. Here these tools are used to gain further insight into scalar transport, and how it is affected by mixing. The aim of this research is to investigate the efficiency of several different partitioning methods which demarcate flow fields into dynamically distinct regions, and the correlation of finite-time statistics from the advection-diffusion equation to these regions. For autonomous systems, invariant manifold theory can be used to separate the system into dynamically distinct regions. Despite there being no equivalent method for nonautonomous systems, a similar analysis can be done. Systems with general time dependencies must resort to using finite-time transport barriers for partitioning; these barriers are the edges of Lagrangian coherent structures (LCS), the analog to the stable and unstable manifolds of invariant manifold theory. Using the coherent structures of a flow to analyze the statistics of trapping, flight, and residence times, the signature of anomalous diffusion are obtained. This research also investigates the use of linear models for approximating the elements of the covariance matrix of nonlinear flows, and then applying the covariance matrix approximation over coherent regions. The first and second-order moments can be used to fully describe an ensemble evolution in linear systems, however there is no direct method for nonlinear systems. The problem is only compounded by the fact that the moments for nonlinear flows typically don't have analytic representations, therefore direct numerical simulations would be needed to obtain the moments throughout the domain. To circumvent these many computations, the nonlinear system is approximated as many linear systems for which analytic expressions for the moments exist. The parameters introduced in the linear models are obtained locally from the nonlinear deformation tensor. / Dissertation/Thesis / Doctoral Dissertation Applied Mathematics 2018

Dados binários ou dicotômicos são comuns em muitas áreas das ciências, nas quais, muitas vezes, há interesse em registrar a ocorrência, ou não, de um evento particular. Por outro lado, quando cada unidade amostral é avaliada em mais de uma ocasião no tempo, tem-se dados longitudinais ou medidas repetidas no tempo. é comum também, nesses estudos, se ter uma ou mais variáveis explicativas associadas às variáveis respostas. As variáveis explicativas podem ser dependentes ou independentes do tempo. Na literatura, há técnicas disponíveis para a modelagem e análise desses dados, sendo os modelos disponíveis extensões dos modelos lineares generalizados. O enfoque do presente trabalho é dado aos modelos lineares generalizados de transição para a análise de dados longitudinais envolvendo uma resposta do tipo binária. Esses modelos são baseados em processos estocásticos e o interesse está em modelar as probabilidades de mudanças ou transições de categorias de respostas dos indivíduos no tempo. A suposição mais utilizada nesses processos é a da propriedade markoviana, a qual condiciona a resposta numa dada ocasião ao estado na ocasião anterior. Assim, são revistos os fundamentos para se especificar tais modelos, distinguindo-se os casos estacionário e não-estacionário. O método da máxima verossimilhança é utilizado para o ajuste dos modelos e estimação das probabilidades. Adicionalmente, apresentam-se testes assintóticos para comparar tratamentos, baseados na razão de chances e na diferença das probabilidades de transição. Outra questão explorada é a combinação do modelo de efeitos aleatórios com a do modelo de transição. Os métodos são ilustrados com um exemplo da área da saúde. Para esses dados, o processo é considerado estacionário de ordem dois e o teste proposto sinaliza diferença estatisticamente significativa a favor do tratamento ativo. Apesar de ser uma abordagem inicial dessa metodologia, verifica-se, que os modelos de transição têm notável aplicabilidade e são fontes para estudos e pesquisas futuras. / Binary or dichotomous data are quite common in many fields of Sciences in which there is an interest in registering the occurrence of a particular event. On the other hand, when each sampled unit is evaluated in more than one occasion, we have longitudinal data or repeated measures over time. It is also common, in longitudinal studies, to have explanatory variables associated to response measures, which can be time dependent or independent. In the literature, there are many approaches to modeling and evaluating these data, where the models are extensions of generalized linear models. This work focus on generalized linear transition models suitable for analyzing longitudinal data with binary response. Such models are based on stochastic processes and we aim to model the probabilities of change or transitions of individual response categories in time. The most used assumption in these processes is the Markov property, in which the response in one occasion depends on the immediately preceding response. Thus we review the fundamentals to specify these models, showing the diferences between stationary and non-stationary processes. The maximum likelihood approach is used in order to fit the models and estimate the probabilities. Furthermore, we show asymptotic tests to compare treatments based on odds ratio and on the diferences of transition probabilities. We also present a combination of random-efects model with transition model. The methods are illustrated with health data. For these data, the process is stationary of order two and the suggested test points to a significant statistical diference in favor of the active treatment. This work is an initial approach to transition models, which have high applicability and are great sources for further studies and researches.

Questões centrais em diversas áreas do conhecimento como ciências ambientais, geologia, agronomia, dentre outras, envolvem a compreensão da distribuição espacial de processos a partir de dados espacialmente referenciados. Os interesses de pesquisa podem estar na descrição espacial de duas ou mais variáveis e, desta forma, tem-se dois ou mais atributos para modelar. Modelos multivariados são propostos para o estudo se há evidências e/ou explicações contextuais de que os processos não são independentes. Diferentes modelos propostos na literatura foram avaliados e comparados ao modelo Matérn multivariado, recentemente proposto na literatura. Foram considerados o modelo linear de corregionalização, o modelo bivariado gaussiano de componente comum e um modelo bayesiano de regressão espacial. Estes modelos foram ajustados e utilizados para predição espacial geoestatística (krigagem) em um conjunto de dados com duas variáveis climáticas no qual uma parte dos dados foi separada para avaliação das predições. Além disso, foi realizado um estudo de simulação para avaliar a estimação e predição sob o modelo Matérn multivariado. / Key issues in a diversity of subject areas such as environmental sciences, geology, agronomy, among other, require the understanding of the spatial distribution of natural processes from spatially referenced data. Research interests may include the spatial description of two or more variables and therefore, there are tow or more attributes to be modeled. Multivariate models are adopted when there is evidence and/or contextual explanations the two processes are not independent. Different models presented in the literature are assessed and compared to the recently introduced multivariate Matérn model. The linear model of corregionalization, the bivariate Gaussian common component model and a bayesian spatial reression model were considered. The models were fitted and used for geostatistical spatial prediction (kriging) for a pair of weather related variables with part of the data used only for comparing the predicions. Additionally a simulation study assessed estimation and prediction under the multivariate Matérn model.

Les modèles linéaires non entiers ont prouvé leur efficacité dans la modélisation de la diffusion thermique pour de faibles variations de température. Cependant, pour de grandes variations de température, les paramètres thermiques dépendent de la température. Par conséquent, la diffusion thermique est régie par un modèle non linéaire non entier. Dans cette thèse, une classe de modèles non linéaires non entiers, basée sur les séries de Volterra non entières, est proposée. Les paramètres non linéaires, tels que les s^n-pôles et l’ordre commensurable, sont estimés au même titre que les coefficients linéaires. Ensuite, le comportement thermique d’un échantillon de fer ARMCO est modélisé pour de grandes variations de température. / Linear fractional differentiation models have proven their efficacy in modeling thermaldiffusive phenomena for small temperature variations. However, for large temperature variations,the thermal parameters are no longer constant but vary along with the temperatureitself. Consequently, the thermal system could be modeled by non linear fractional differentialmodels. Volterra series are first extended to fractional derivatives. Volterra seriesare then used for modeling a non linear thermal system, constituted of an ARMCO iron sample,for large temperature variations.

This thesis discusses the changes in parameters of linear models and methods of their detection. It begins with a short introduction of the two basic types of change point detection procedures and bootstrap algorithms developed specifically to deal with dependent data. In the following chapter we focus on the location model - the simplest example of a linear model with a change in parameters. On this model we will illustrate a way of long-run variance estimation and implementation of selected bootstrap procedures. In the last chapter we show how to extend the applied methods to linear models with a change in parameters. We will compare the performance of change point tests based on asymptotic and bootstrap critical values through simulation studies in both our considered methods. The performance of selected long-run variance estimator will also be examined both for situations when the change in parameters occurs and when it does not. 1

Les données à haut-débit, par leur grande dimension et leur hétérogénéité, ont motivé le développement de méthodes statistiques pour la sélection de variables. En effet, le signal est souvent observé simultanément à plusieurs facteurs de confusion. Les approches de sélection habituelles, construites sous l'hypothèse d'indépendance des variables, sont alors remises en question car elles peuvent conduire à des décisions erronées. L'objectif de cette thèse est de contribuer à l'amélioration des méthodes de sélection de variables pour la régression et la classification supervisée, par une meilleure prise en compte de la dépendance entre les statistiques de sélection. L'ensemble des méthodes proposées s'appuie sur la description de la dépendance entre covariables par un petit nombre de variables latentes. Ce modèle à facteurs suppose que les covariables sont indépendantes conditionnellement à un vecteur de facteurs latents. Une partie de ce travail de thèse porte sur l'analyse de données de potentiels évoqués (ERP). Les ERP sont utilisés pour décrire par électro-encéphalographie l'évolution temporelle de l'activité cérébrale. Sur les courts intervalles de temps durant lesquels les variations d'ERPs peuvent être liées à des conditions expérimentales, le signal psychologique est faible, au regard de la forte variabilité inter-individuelle des courbes ERP. En effet, ces données sont caractérisées par une structure de dépendance temporelle forte et complexe. L'analyse statistique de ces données revient à tester pour chaque instant un lien entre l'activité cérébrale et des conditions expérimentales. Une méthode de décorrélation des statistiques de test est proposée, basée sur la modélisation jointe du signal et de la dépendance à partir d'une connaissance préalable d'instants où le signal est nul. Ensuite, l'apport du modèle à facteurs dans le cadre général de l'Analyse Discriminante Linéaire est étudié. On démontre que la règle linéaire de classification optimale conditionnelle aux facteurs latents est plus performante que la règle non-conditionnelle. Un algorithme de type EM pour l'estimation des paramètres du modèle est proposé. La méthode de décorrélation des données ainsi définie est compatible avec un objectif de prédiction. Enfin, on aborde de manière plus formelle les problématiques de détection et d'identification de signal en situation de dépendance. On s'intéresse plus particulièrement au Higher Criticism (HC), défini sous l'hypothèse d'un signal rare de faible amplitude et sous l'indépendance. Il est montré dans la littérature que cette méthode atteint des bornes théoriques de détection. Les propriétés du HC en situation de dépendance sont étudiées et les bornes de détectabilité et d'estimabilité sont étendues à des situations arbitrairement complexes de dépendance. Dans le cadre de l'identification de signal, une adaptation de la méthode Higher Criticism Thresholding par décorrélation par les innovations est proposée. / The analysis of high throughput data has renewed the statistical methodology for feature selection. Such data are both characterized by their high dimension and their heterogeneity, as the true signal and several confusing factors are often observed at the same time. In such a framework, the usual statistical approaches are questioned and can lead to misleading decisions as they are initially designed under independence assumption among variables. The goal of this thesis is to contribute to the improvement of variable selection methods in regression and supervised classification issues, by accounting for the dependence between selection statistics. All the methods proposed in this thesis are based on a factor model of covariates, which assumes that variables are conditionally independent given a vector of latent variables. A part of this thesis focuses on the analysis of event-related potentials data (ERP). ERPs are now widely collected in psychological research to determine the time courses of mental events. In the significant analysis of the relationships between event-related potentials and experimental covariates, the psychological signal is often both rare, since it only occurs on short intervals and weak, regarding the huge between-subject variability of ERP curves. Indeed, this data is characterized by a temporal dependence pattern both strong and complex. Moreover, studying the effect of experimental condition on brain activity for each instant is a multiple testing issue. We propose to decorrelate the test statistics by a joint modeling of the signal and time-dependence among test statistics from a prior knowledge of time points during which the signal is null. Second, an extension of decorrelation methods is proposed in order to handle a variable selection issue in the linear supervised classification models framework. The contribution of factor model assumption in the general framework of Linear Discriminant Analysis is studied. It is shown that the optimal linear classification rule conditionally to these factors is more efficient than the non-conditional rule. Next, an Expectation-Maximization algorithm for the estimation of the model parameters is proposed. This method of data decorrelation is compatible with a prediction purpose. At last, the issues of detection and identification of a signal when features are dependent are addressed more analytically. We focus on the Higher Criticism (HC) procedure, defined under the assumptions of a sparse signal of low amplitude and independence among tests. It is shown in the literature that this method reaches theoretical bounds of detection. Properties of HC under dependence are studied and the bounds of detectability and estimability are extended to arbitrarily complex situations of dependence. Finally, in the context of signal identification, an extension of Higher Criticism Thresholding based on innovations is proposed.

The final thesis deals with social situations juniors and seniors in selected countries of the European Union. The thesis monitors changes in social developments primarily in terms of income, education, and especially of unemployment. The selected period is the period from approximately 2005 to 2011, in some cases up to 2013. The aim of this thesis is the statistical analysis of the data set EU-SILC for 2005 and 2010 of Czech Republic, Slovakia, Poland, Germany, France and Spain with focusing on income, education and unemployment among age groups.

Neste trabalho estudamos as aplicações do método do Gradiente Espectral Projetado (SPG) em Meteorologia nos campos de previsibilidade, estabilidade e sensibilidade. Inicialmente revisamos os Vetores Singulares Lineares (LSVs) e em seguida apresentamos a teoria das Perturbações Condicionais Não-Lineares Ótimas (CNOPs). Enquanto os métodos clássicos estão baseados no Modelo Tangente Linear, as CNOPs são uma formulação do mesmo problema baseado em Programação Não-Linear. As CNOPs são descritas na literatura como responsáveis por melhorias em relação aos métodos anteriores. Finalmente analisamos três exemplos de aplicação do método à problemas de previsibilidade, estabilidade e sensibilidade. / A revision about applications of Spectral Projected Gradient (SPG) in meteorology is done in the fields of predictability, stability and sensitivity. Initially we review about Linear Singular Vectos (LSVs) and we present the Conditional Nonlinear Optimal perturbations (CNOPs). While the classic methods are based on the Tangent Linear Model, CNOPs are another formulation of the problem based on Nonlinear Programming. CNOPs are described in bibliography as responsible by better results than older methods. Finally we analyze three applications in predictability, stability and sensibility.