This dissertation focuses on trends in pay inequality in the formal sectors of Argentina, Brazil, and Chile from the early 1990s into the latter part of the first decade of the new millennium. In-depth, single-country studies of inequality of each country of study seek to understand and explain the sources of movement in inequality in each country, relating changes in inequality to shifts in the relative roles of key economic sectors and geographic jurisdictions. In addition to these single-country studies of inequality, this dissertation develops a regional perspective on the dynamics of inequality by synthesizing findings from the three countries of study, identifying both commonalities and differences. This dissertation also evaluates the relationship between trends in inequality and the macroeconomic policies and factors that influence them. By eschewing the inequality of household incomes and focusing instead on measures of inequality in the underlying distribution of pay, this dissertation presents empirical evidence that fluctuations in countries' inequality levels are intrinsically related to macroeconomic factors. This dissertation applies Theil's T statistic, which belongs to the family of generalized entropy inequality measures, to develop new measures of economic inequality. The calculations presented in this dissertation are performed on data obtained from semi-aggregated datasets in which employment and average wage data organized by economic sectors and geographical jurisdictions, as derived from administrative records. Sectoral analysis shows that the changing levels of overall inequality are explained to a great extent by variations in the performance of a reduced number of "key" high-pay sectors, especially finance, extractive industry and civil service. In terms of the dynamics of geographic distribution, the role of these key sectors is observed in the driving role played by key geographic units: those composed of, or containing, the countries' main metropolitan centers, and those with high concentrations of economic activity in extractive industries. / text

We first show theoretically and in simulation how power varies as a function of SNP correlation structure with currently-implemented gene-based testing methods. We propose alternative testing methods whose power does not vary with the correlation structure. We then propose hypothesis tests for detecting prevalence-incidence bias in case-control studies, a bias perhaps overrepresented in GWAS due to currently used study designs. Lastly, we hypothesize how different incentive structures used to keep clinical trial participants in studies may interact with a background of dependent censoring and result in variation in the bias of the Kaplan-Meier survival curve estimator.

Šiame magistro darbe nagrinėjama IR statistika bei FARIMA ir AR duomenų generavimo algoritmai. Darbe pateikta programinė IR statistikos skaičiavimo bei FARIMA ir AR duomenų generavimo algoritmų realizacija. Visi algoritmai visapusiškai ištestuoti bei jiems sukurta vartotojo sąsaja. IR statistika dar nėra pilnai įrodyta, todėl šis darbas galės būti naudojamas tolimesniems IR statistikos tyrimams. / IR statistics, FARIMA and AR data production algorithms are analyzed in this master’s degree study. I present IR statistics calculating, FARIMA and AR data production algorithms software realization. All algorithms are universally tested and has user interface. IR statistics is not fully proven, that’s why this study could be used in further IR statistics researches.

In this thesis, we study two methods for inference of parameters in the accelerated failure time model with right censoring data. One is the Wald-type method, which involves parameter estimation. The other one is empirical likelihood method, which is based on the asymptotic distribution of likelihood ratio. We employ a monotone censored data version of Kendall estimating equation, and construct confidence intervals from both methods. In the simulation studies, we compare the empirical likelihood (EL) and the Wald-type procedure in terms of coverage accuracy and average length of confidence intervals. It is concluded that the empirical likelihood method has a better performance. We also compare the EL for Kendall’s rank regression estimator with the EL for other well known estimators and find advantages of the EL for Kendall estimator for small size sample. Finally, a real clinical trial data is used for the purpose of illustration.

In order to determine the profile of pupils and the Militar Schools, multivariate statistical techniques on the students data are applied, providing therefore, aids for the administration decision takings. First, a data descriptive analysis of four Militar Schools was carried out according to a cross of school efficiency variants. The multivariate analysis is applied in some teaching indicators by using agglomeration analysis as well as principal components and discriminatory. It is possible to identify the number of students presenting low school grades in their social classes and to compare them among the schools. Once grades in disciplines are used, sets that represent attributes in psychomotor/affective and scientific/cognitive areas are verified. A model formed by the most representative disciplines in the school year, compared to the need of recovery, is shown, classifying a new student in this system. The techniques applied in teaching indicators seem to be suitable to check the quality because when the multivariate nature is obeyed, relevant information can be taken by using different variants and cases. The search for knowledge in data banks has been distinguished as a desirable activity to find out and to understand hidden standards in institutions once they are not always seen through ordinary observation. Thus, it is aimed to enhance the competence and creativity in public institutions looking for organization and quality system management by the usage of methodology in order to show the comparative performance among schools and their own pupils / Com o objetivo de determinar o perfil dos alunos e dos Colégios Militares, aplicam-se técnicas estatísticas multivariadas em dados dos alunos, disponibilizando assim subsídios para a tomada de decisões da administração. Realiza-se, primeiramente, a análise descritiva dos dados de quatro Colégios Militares, segundo um cruzamento de variáveis de rendimento escolar. Aplica-se análise multivariada em alguns indicadores de ensino, utilizando-se análise de aglomeração, componentes principais e discriminante. Pode-se identificar o número de alunos com rendimento baixo nas classes de origem social e compará-los entre as escolas. Utilizando graus das disciplinas, verificam-se agrupamentos que representam os atributos da área psicomotora/afetiva e ciências/cognitivas. Mostra-se um modelo formado pelas disciplinas mais representativas em um ano letivo comparadas com a necessidade de recuperação, classificando um aluno novo neste sistema. As técnicas aplicadas em indicadores de ensino mostram-se adequadas para a verificação da qualidade, pois, obedecendo a natureza multivariada, pode-se extrair informações relevantes utilizando-se diferentes casos e variáveis. A procura de conhecimento em bancos de dados tem se destacado como uma atividade desejável para descobrir e compreender padrões ocultos nas instituições, os quais nem sempre são visíveis através da simples observação. Desta forma, busca-se aumentar a competência e a criatividade nas instituições públicas, visando a organização e gestão de sistemas de qualidade, através do uso de metodologia para mostrar o desempenho comparativo entre as escolas e entre os próprios alunos.

Ce travail de thèse porte sur une meilleure compréhension de la croissance de nanoparticules d'oxyde de zinc élaborées à partir d'une approche organométallique. Au cours de ce travail, une analyse de la taille des nanoparticules en utilisant un outil simple basée sur une représentation bidimentionnelle couplée à une analyse statistique multivariée. Elle permet d'extraire la corrélation éventuelle entre la longueur et la largeur des nanoparticules. Cette approche permet de traiter un ensemble de nanoparticules homogène ou hétérogène en taille et en forme. Comparée à l'approche statistique usuelle qui analyse individuellement la distribution de la longueur et celle de la largeur, notre démarche met très facilement en évidence les différents types de nanoparticule et leur anisotropie (ou non). Pour chaque classe de nano-objets considérée, la relation entre la longueur et la largeur, c'est-à-dire les corrélations fortes ou faible entre ces deux paramètres, fourni des informations concernant le(s) mécanisme(s) de nucléation-croissance des nanoparticules. Par exemple, cette approche permet de suivre l'effet sur la taille et la forme de nanoparticules préparées par différents procédés (chimiques ou physique). A partir de différentes images de microscopie électronique extraites de la bibliographie ou de nos travaux de recherche, nous avons démontré l'efficacité et la simplicité de cette approche. De plus, nous avons pu revisiter la synthèse organométallique de nanoparticules de ZnO à la lumière de cette nouvelle méthode de traitement des données. Les nanoparticules anisotropes obtenues par hydrolyse du dicyclohexyl zinc comme précurseur organométallique en présence d'alkyl amine uniquement ont donc été analysées. Le mécanisme de formation a été étudié par analyse systématique des images de microscopie électroniques à transmission et une croissance par attachement orienté a pu être mise en évidence. De plus, nous avons montré que la taille des nanabatonnets d'oxyde de zinc peut être facilement contrôlée en modulant la vitesse d'hydrolyse, le temps de maturation du mélange précurseur organométallique et ligand alkyl amine avant le début de l'hydrolyse, la nature du ligand et la quantité d'eau introduite. Enfin, pendant ce travail, nous avons observé la gélification des milieux réactionnels. L'étude de ce phénomène a permis de développer une approche originale permettant de mettre en forme des matériaux hybride organique-inorganiques. En effet, à partir d'un ensemble d'expériences utilisant la rhéologie, la RMN, la MET, des calculs théoriques, le WAXS ... nous avons pu mettre en évidence une réaction d'oligomérisation entre le précurseur organométallique et les ligands alkyl amine. Cette oligomérisation entre les centres métalliques conduit à la formation de gels, qui peuvent être mise en forme en utilisant des seringues, des tampons ou encore des moules. Des matériaux hybrides d'oxydes métalliques sont ensuite obtenus par hydrolyse du gel. Cette nouvelle approche a été développée pour l'oxyde de zinc puis a été étendue aux oxydes d'étain et de fer. Il s'agit donc d'une approche générale qui ouvre des perspectives intéressantes dans l'utilisation de ces matériaux notamment comme couches sensibles pour des capteurs de gaz. / This work is focused on the better understanding of the formation of zinc oxide anisotropic nanoparticle through an organometallic synthesis. To start on this project, the development of a statistical tool to analyze nanoparticle sizes and shapes was mandatory. A simple 2D plot with a multivariate statistical analysis is proposed in order to extract, in particular, the correlation between length and width in a collection or a mixture of anisotropic particles. Compared to the usual statistics on the length associated to a second and independent statistical analysis of the width, this simple plot easily points out the various types of nanoparticles and their (an)isotropy. For each class of nano-objects, the relationship between width and length (i.e. the strong or weak correlations between these two parameters) may suggest information concerning the nucleation-growth processes. It allows one to follow the effect on the shape and size distribution of physical or chemical processes such as simple ripening. Various electronic microscopy pictures from the literature or from our own syntheses are used as examples to demonstrate the efficiency and simplicity of such an analysis. Furthermore, following the organometallic approach, well defined crystalline zinc oxide nanorods are obtained through the hydrolysis reaction of dicyclohexyl zinc precursor without solvent in presence of primary fatty amine only. Their mechanism of formation is studied through a multivariate analysis of TEM results and an oriented attachment process is demonstrated. Taking advantage of this mechanism, the size of the ZnO nanorods can be easily adjusted tuned by controlling adjusting the hydrolysis rate, the aging time before hydrolysis, the nature of ligand, and/or the amount of water. During such experiments, a gelification of the chemical media are for the first time observed. This is the origin of the last part of our work, which put forward an original approach for outstanding processable and reshapable hybrid materials. Indeed, we unequivocally demonstrate through various complementary experiments (rheology, NMR, TEM, theoretical calculation, WAXS...) the formation of oligomeric structures in a mixture of fatty amine with organometallic compounds. This oligomerization reaction between the metallic centers induces the formation of a gel which can be processed into fibers, stamped patterns or molded pieces. Metal oxide hybrid materials are obtained merely by hydrolysis of the organometallic gel, which are furthermore liquid crystal. This unprecedented route towards easily processable hydrid and metal oxide materials is demonstrated for zinc oxide and extended to other metal oxides such as tin oxide and iron oxide, proving the universality of the method and paving the way, for example, to new sensitive gas sensors layers.

L'accroissement rapide des données numériques vidéographiques fait de la compréhension automatiquedes vidéos un enjeu de plus en plus important. Comprendre de manière automatique une vidéo recouvrede nombreuses applications, parmi lesquelles l'analyse du contenu vidéo sur le web, les véhicules autonomes,les interfaces homme-machine. Cette thèse présente des contributions dans deux problèmes majeurs pourla compréhension automatique des vidéos : la détection d'actions supervisée par des données web, et la localisation d'actions humaines.La détection d'actions supervisées par des données web a pour objectif d'apprendre à reconnaître des actions dans des contenus vidéos sur Internet, sans aucune autre supervision. Nous proposons une approche originaledans ce contexte, qui s'appuie sur la synergie entre les données visuelles (les vidéos) et leur description textuelle associée, et ce dans le but d'apprendre des classifieurs pour les événements sans aucune supervision. Plus précisément, nous télechargeons dans un premier temps une base de données vidéos à partir de requêtes construites automatiquement en s'appuyant sur la description textuelle des événéments, puis nous enlevons les vidéos téléchargées pour un événement, et dans laquelle celui-ci n'apparaït pas. Enfin, un classifieur est appris pour chaque événement. Nous montrons l'importance des deux étapes principales, c'est-à-dire la créations des requêtes et l'étape de suppression des vidéos, par des résutatsquantitatifs. Notre approche est évaluée dans des conditions difficiles, où aucune annotation manuelle n'est disponible, dénotées EK0 dans les challenges TrecVid. Nous obtenons l'état de l'art sur les bases de donnéesMED 2011 et 2013.Dans la seconde partie de notre thèse, nous nous concentrons sur la localisation des actions humaines, ce qui implique de reconnaïtre à la fois les actions se déroulant dans la vidéo, comme par exemple "boire" ou "téléphoner", et leur étendues spatio-temporelles. Nous proposons une nouvelle méthode centrée sur la personne, traquant celle-ci dans les vidéos pour en extraire des tubes encadrant le corps entier, même en cas d'occultations ou dissimulations partielles. Deux raisons motivent notre approche. La première est qu'elle permet de gérer les occultations et les changements de points de vue de la caméra durant l'étape de localisation des personnes, car celle-ci estime la position du corps entier à chaque frame. La seconde est que notre approche fournit une meilleure grille de référence que les tubes humains standards (c'est-à-dire les tubes qui n'encadrent que les parties visibles) pour extraire de l'information sur l'action. Le coeur de notre méthode est un réseau de neurones convolutionnel qui apprend à générer des propositions de parties du corps humain. Notre algorithme de tracking connecte les détections temporellement pour extraire des tubes encadrant le corps entier. Nous évaluons notre nouvelle méthode d'extraction de tubes sur une base de données difficile, DALY, et atteignons l'état de l'art. / With the rapid growth of digital video content, automaticvideo understanding has become an increasingly important task. Video understanding spansseveral applications such as web-video content analysis, autonomous vehicles, human-machine interfaces (eg, Kinect). This thesismakes contributions addressing two major problems in video understanding:webly-supervised action detection and human action localization.Webly-supervised action recognition aims to learn actions from video content on the internet, with no additional supervision. We propose a novel approach in this context, which leverages thesynergy between visual video data and the associated textual metadata, to learnevent classifiers with no manual annotations. Specifically, we first collect avideo dataset with queries constructed automatically from textual descriptionof events, prune irrelevant videos with text and video data, and then learn thecorresponding event classifiers. We show the importance of both the main steps of our method, ie,query generation and data pruning, with quantitative results. We evaluate this approach in the challengingsetting where no manually annotated training set is available, i.e., EK0 in theTrecVid challenge, and show state-of-the-art results on MED 2011 and 2013datasets.In the second part of the thesis, we focus on human action localization, which involves recognizing actions that occur in a video, such as ``drinking'' or ``phoning'', as well as their spatial andtemporal extent. We propose a new person-centric framework for action localization that trackspeople in videos and extracts full-body human tubes, i.e., spatio-temporalregions localizing actions, even in the case of occlusions or truncations.The motivation is two-fold. First, it allows us to handle occlusions and camera viewpoint changes when localizing people, as it infers full-body localization. Second, it provides a better reference grid for extracting action information than standard human tubes, ie, tubes which frame visible parts only.This is achieved by training a novel human part detector that scores visibleparts while regressing full-body bounding boxes, even when they lie outside the frame. The core of our method is aconvolutional neural network which learns part proposals specific to certainbody parts. These are then combined to detect people robustly in each frame.Our tracking algorithm connects the image detections temporally to extractfull-body human tubes. We evaluate our new tube extraction method on a recentchallenging dataset, DALY, showing state-of-the-art results.

Esta Dissertação de Mestrado tem como objetivo analisar como os discursos que circulam nas páginas da Ação ABC Alfabetizando se constituem em estratégias de governamento e operam sobre os governos municipais e as população para manter os índices de analfabetismo reduzidos. A Ação ABC Alfabetizando é uma iniciativa nãogovernamental, que se institui visando a erradicação do analfabetismo nas regiões do Vale dos Sinos, Vale do Caí e Vale do Paranhana & Encosta da Serra, compreendendo 44 municípios do Rio Grande do Sul. Analiso os discursos coletados a partir das páginas semanais da Ação, compreendidas entre 2004 a 2008. As escolhas metodológicas inserem-se na Linha de Pesquisa dos Estudos Culturais em Educação, em sua vertente pós-estruturalista e valem-se dos estudos sobre governamentalidade e sobre o discurso, conforme tematizados por Michel Foucault. Destaco, também, trabalhos dos autores Clarice Salete Traversini, Iole Trindade, Afredo Veiga-Neto e Rosa Maria Fischer Bueno que auxiliaram na realização das análises. Com a analítica desenvolvida, percebi que os discursos veiculados nas páginas da Ação ABC Alfabetizando se constituem em estratégias de governamento a partir de mitos, da mídia e da estatística. Os discursos dos mitos do alfabetismo-como-uma-via-para-odesenvolviment e do alfabetismo-como-uma-via-para-a-cidadania, atribuem ao alfabetismo a condição da decolagem ao sucesso, ao desenvolvimento social e ao exercício da cidadania, enquanto que ao espírito comunitário é creditado o esforço coletivo para o sucesso da alfabetização. A mídia é o espaço utilizado pela Ação, onde são vizibilizadas as ações desenvolvidas nos municípios. O saber estatístico se constitui numa estratégia da Ação, que opera em três movimentos: conhecer (os dados anteriores à Ação), intervir ( a partir dos dados, implantar programas de alfabetização) e avaliar (resultados). Esses discursos circulam e agem, operando sobre a população na condução das suas condutas em favor da alfabetização, reduzindo, assim, os índices de analfabetismo na região. / This Master dissertation is aimed to analyze how the speeches present in the pages from ‘Ação ABC Alfabetizando’ (ABC Alphabetization Action) constitute strategies of government and act on municipal government and the population to keep illiteracy rates low. The ‘Ação ABC Alfabetizando’ is a non-governmental initiative and its goal is to eradicate illiteracy in the regions of Vale dos Sinos, Vale do Caí , Vale do Paranhana and Encosta da Serra, formed by 44 cities in Rio Grande do Sul. I analyze the speeches collected from the weekly pages of ‘Ação’, between 2004 and 2008. The methodological choices are inserted in the ‘Linha de Pesquisa dos Estudos Culturais em Educação’ (Cultural Studies Research in Education) in the post-structuralism line and are related to the studies about how to govern and the speech, as according to Michel Foucault. I also emphasize the work from the authors Clarice Salete Traversini, Iole Trindade, Afredo Veiga-Neto and Rosa Maria Fischer Bueno that helped in the analysis. With the development, I realized that the speeches present in the pages of ‘Ação ABC Alfabetizando’ are constituted in strategies of government from myths, media and statistics. The speeches of the myths of the literacy as a way of development and citizenship confer to it the condition to take-off to success, to social development and to citizen practice, while the collective effort to the success of the alphabetization is related to the communitarian spirit. The media is the space used by the ‘Ação’, in which the actions developed in the cities are feasible. The statistic knowledge is a strategy of ‘Ação’, present in three moves: to know (data before the Action), to intervene (taking data into consideration, to implant alphabetization programs) and to evaluate (results). These speeches surround and act, influencing the population in their conducts and their practices in favor of the alphabetization, thus reducing the illiteracy rates in the region.

O concreto é o segundo material mais consumido no mundo, perdendo somente para água, justificando a importância de estudos que envolvam seu comportamento frente à durabilidade, e que objetivam a obtenção de estruturas com longas vidas úteis. Dentre as diversas manifestações patológicas que ocorrem nas estruturas de concreto armado, a corrosão das armaduras tem uma grande incidência, como já comprovado por diversos autores. A corrosão das armaduras ocorre pela ação de agentes agressivos sendo que os dois principais são a carbonatação e a penetração de cloretos. Em função da grande costa marítima do Brasil, a ação de íons cloretos é a mais significativa, deteriorando estruturas e levando a necessidade de manutenções periódicas. Porém, sabe-se do alto custo para realização destas manutenções. Em função da preocupação com a manutenção e gastos das estruturas de concreto armado, este estudo objetivou realizar uma validação dos ensaios de corrosão por cloretos em estruturas de concreto armado a partir de meta-análise dos dados de técnicas eletroquímicas utilizadas nos trabalhos de teses e dissertações desenvolvidos no Brasil. Pretende-se desta maneira realizar um mapeamento das pesquisas na área de corrosão em armaduras de concreto armado realizadas no Brasil e direcionar os futuros estudos. Pretende-se também realizar uma análise conjunta de todos estes trabalhos para verificar a quantidade e qualidade dos mesmos sob uma visão do conjunto nacional. Registrou-se a falta de pesquisas na região Norte do país e a centralização das mesmas nas regiões Sudeste e Sul. Com o uso de técnicas estatísticas conclui-se a falta de diversas combinações de variáveis (relação água/cimento, tipo de cimento, método de indução de cloretos, adição de materiais, entre outros) para criação de um modelo brasileiro confiável. / Concrete is the second most consumed material in the world, second only to water, explaining the importance of studies involving behavior concerning durability, and aimed to obtain structures with long lifetimes. Among the various pathologics that occur in reinforced concrete structures, corrosion of the reinforcement has a major impact, as proven by several authors. The reinforcement corrosion occurs by the action of aggressive agents and the two main ones are carbonation and chloride penetration. Due to the great sea coast of Brazil, the action of chloride ions is the most significant and deteriorating structures leading to the need for periodic maintenance, but it is known the high cost of implementing these maintenances. Because of concerns with the maintenance and expense of concrete structures, this study aimed to perform a validation of the corrosion by chlorides in reinforced concrete structures from the meta-analysis of data from electrochemical techniques used in the work of theses and dissertations developed in Brazil. The aim is thus to map the research on corrosion in reinforced concrete made in Brazil and direct future studies. Also perform a joint analysis of all these works to verify the quantity and quality of these in a vision of the National Assembly. We observed the lack of research in the northern region of the country and centralizing them in the Southeast and South By using statistical techniques concludes the lack of various combinations of variables (water / cement ratio, cement type, method induction of chloride, addition of materials, among others) to create a Brazilian model reliable.

La protection de l’environnement, en particulier celle des systèmes aquatiques, est une des priorités de nos sociétés. L’utilisation de capteurs biologiques permettant de tester la qualité de l’eau en continue est une voie possible de surveillance intégrée des milieux aquatiques. Cette démarche a été mise en place avec succès sur des mollusques bivalves équipés d’électrodes légères qui respectent leur comportement naturel, on parle alors de valvométrie. Le but de cette thèse est de calculer et traiter automatiquement la vitesse de mouvement des valves de mollusques bivalves installés dans divers milieux aquatiques. Les années d’enregistrements déjà acquises nous permettrons, à partir de nos modèles, de détecter s’il existe des variations de la vitesse de mouvement des valves liées aux variations de température. Plus particulièrement, nous avons étudié les dérivées de différents estimateurs non paramétriques d’une fonction de régression : l’estimateur récursif de Nadaraya-Watson, l’estimateur de Johnston, l’estimateur de Wand-Jones ainsi que l’estimateur de Révész. Nous avons aussi pris en compte la version déterministe de l’estimateur de Nadaraya-Watson. Pour chacun des estimateurs nous avons mené une étude sur les comportement asymptotiques en particulier la convergence presque sûre et la normalité asymptotique. Nous avons illustré numériquement ces propriétés et appliqué ces nouvelles méthodes d’estimations sur des données réelles afin de valider, ou non, les hypothèses environnementales émises par les biologistes. / The protection of the environment, in particular aquatic systems, should be tackled as one of the top priorities of our society. The use of biological sensors to continuously test water quality is a possible way of monitoring aquatic environments. This approach has been successfully implemented on bivalve molluscs equipped with light electrodes that respect their natural behaviour, we then speak of valvometry. The purpose of this thesis is to automatically calculate and process the velocity of movement of bivalve mollusc valves in various aquatic environments. Years of recordings already acquired will allow us, from our models, to detect if there are variations in the speed of movement of the valves related to temperature variations. In particular, we studied the derivatives of different non parametric estimators of the regression function : the recursive Nadaraya-Watson estimator, the Johnston estimator, the Wand-Jones estimator and the Révész estimator. We also considered the deterministic version of the Nadaraya-Watson estimator. For each of the estimators we conducted a study on the asymptotic behaviour especially on the almost sure convergence and th asymptotic normality. We digitally illustrated these properties and applied these new estimation methods to real data to validate, or not, the environmental assumptions made by biologists.