Estudos genético quantitativos e moleculares são fundamentais para o melhoramento animal e sua realização com a raça Nelore é de grande importância devido a ampla participação dessa no rebanho de corte nacional. A estimação constante dos parâmetros genéticos das características de produção é necessário para a adequada condução do processo de seleção dos animais. A melhoria de características relacionadas à carcaça bovina é essencial para a eficiência e sustentabilidade da atividade e a implementação de métodos de seleção animal baseados em informações moleculares pode revolucionar a produção zootécnica e deve ser profundamente estudado. Sendo assim, os objetivos do presente estudo foram estimar parâmetros genéticos e componentes de variância através de diferentes modelos matemáticos para um total de 14 características fenotípicas (o peso ao nascimento, peso a desmama, peso ao sobreano, ganho de peso entre a desmama e o sobreano ajustado para um intervalo de 345 dias, perímetro escrotal ao sobreano, altura de garupa ao sobreano, escores visuais avaliados ao sobreano de conformação, precocidade, musculosidade, comprimento de umbigo e ossatura, e ainda características de carcaça mensuradas por ultrassonografia realizada após 30 a 45 dias de confinamento como a área de olho de lombo, espessura de gordura subcutânea, espessura de gordura na picanha). Foram estimadas correlações entre todas estas características com as de carcaça mensuradas por ultrassonografia. Sob o enfoque molecular, desenvolveu-se um programa para imputação de genótipos faltantes e estudaram-se diferentes métodos de associação de marcadores moleculares do tipo mutação de base nitrogenada única (SNP) a características de produção incluídas no índice de seleção de um programa de melhoramento da raça Nelore, utilizando inferência bayesiana. Todas as características estudadas podem ser selecionadas esperando-se progresso genético na população. Os efeitos maternos foram importantes em algumas características onde normalmente estes efeitos não têm sido considerados atualmente. A quantidade de escores atribuídos a uma característica categórica assim como o número de observações fenotípicas resultam em diferenças nas estimativas quando avaliadas por modelos lineares ou de limiar. Não deverão ser obtidos resultados satisfatórios na melhoria das características de carcaça se a seleção for baseada nas tradicionais avaliações visuais utilizadas no momento. Os métodos utilizados na análise de associação dos marcadores podem originar diferentes resultados. Os marcadores que apresentaram efeitos altamente relevantes (P<0,01) geralmente apresentaram resultados semelhantes, independentemente do método utilizado. Certos marcadores podem ter efeitos positivos para algumas características componentes do índice de seleção e negativos para as demais. A análise em conjunto com todos os SNP\'s e todos os dados fenotípicos disponíveis é viável e parece ser a mais adequada. O método desenvolvido de imputação de genótipos faltantes a partir do parentesco de animais genotipados foi eficiente. / Quantitative and molecular genetic studies are very important for animal breeding and studies with Nellore cattle have great importance due to the large participation of that breed in the Brazilian beef cattle industry (around 80% of the herd). The constant estimation of genetic parameters for traits linked to production is necessary for properly perform selection of animals. The improvement of carcass traits is essential for efficiency and profitability of the activity. The implementation of methods for animal selection based on molecular information could revolutionize animal production and should be deeply studied. Thus, the objectives of this study were to estimate genetic parameters and variance components using different mathematical models for a total of 14 traits, such as birth weight, weaning weight, yearling weight, post-weaning weight gain between weaning and yearling adjusted for 345 days, yearling scrotal circumference, yearling hip height, yearling visual scores like conformation, finishing, muscularity, bone structure and navel length. Ultrasound measurements for carcass traits performed at feedlot (30 to 45 days, at approximate age of 20 months) such as rib-eye area, fat thickness, rump fat thickness were, also, evaluated. Estimate correlations between all these traits with the carcass traits measured by ultrasound were estimated. As concerned to molecular study, an algorithm for imputation of missing genotypes was developed and different methods to analyze molecular marker (single nucleotide polymorphism - SNP) association with traits components of the selection index of a breeding program that is applied to the population studied, using bayesian inference, were used. Genetic progress will be expected for selection of all the traits studied. Maternal effects were important in some traits in which those effects are not usually considered. The amount of scores assigned to a categorical trait and the number of observations could result in different estimates when evaluated by linear or threshold models. The selection for visual scores traditionally used in that population will not improve carcass traits. The methods used to analyze the association of markers may lead to different results. The SNP\'s with association effects of high relevance (P<0.01) generally express their effects regardless of the method used to analyze. Some markers may have positive effects for some traits of selection index, but negative for others. The joint analysis with all SNPs and with all available phenotypes is feasible and appears to be more appropriate. The algorithm developed for imputation of missing genotypes from pedigree information of genotyped animals was efficient.

本論文探討不同的物價環境對匯率轉嫁程度的影響。有別於既有文獻將通貨緊縮納入低通膨的區間一併討論，本論文特別區分通貨緊縮與低且正的通膨區間，估計不同物價環境下的匯率轉嫁效果。 利用門檻迴歸模型（threshold model）及台灣1981-2008年的資料，且區分能源價格及非能源價格，本文的實證結果顯示，通貨緊縮考慮與否將影響匯率轉嫁程度與通膨、通縮環境的關係。不同於既有文獻發現匯率轉嫁效果與通膨環境呈正相關，本文發現通貨緊縮環境下的匯率轉嫁效果會提高。此外，包含能源價格之匯率轉嫁效果隨物價環境變化的幅度較大，與既有文獻的看法一致。 因此，在匯率轉嫁效果與物價環境的分析上，明確區分通貨緊縮的情況有其必要性，否則可能形成偏誤之推論。 / This dissertation incorporates inflation and deflation in the analysis of exchange rate pass-through at different price levels. Because the existing literature generally consider deflation as part of low inflation, pass-through estimates tend to be considered the same for these two regimes. This study separates the effects of deflation and low positive inflation and estimates the pass-through for different price levels. This dissertation uses a nonlinear model with aggregate and disaggregate import prices data from 1981–2008 in Taiwan to first examine the pass-through for two regimes of high inflation and low inflation. The results confirm the notion in the literature that a positive relationship exists between pass-through and inflation. Then, this dissertation extends the model to a three-regime setting, including high inflation, low positive inflation, and deflation. When deflation is clearly defined in a three-regime model, the degree of exchange rate pass-through is found to be increasing in both high inflation and deflation. The positive relationship at all price levels is no longer valid while the effect of deflation is separated from that of low inflation. In Taiwan, the pass-through becomes inversely greater as the inflation rate falls into a deflationary regime. That the pass-through is higher in a deflationary regime became particularly obvious after the 1997 financial crisis. Contrary to the results predicted by the positive relationship, this analysis does not find an unlimited downward trend for the pass-through. A rebound occurs in the degree of pass-through once deflation is clearly identified, and this pattern is also found for half of the importing industries categorized using the Standard International Trade Classification (SITC). In addition, the results are consistent with the notion that oil prices usually fluctuate much more than the prices of other imports. The estimates show that the pass-through changes the most for fuels and related materials. Obviously, fluctuations in the price of oil influence the measurement of the pass-through. The increase in the pass-through found in a deflationary regime becomes smaller when oil prices are excluded.

Popularity of nonlinear threshold models and unit root tests has increased after the recent empirical studies concerning the effects of business cycles on macroeconomic data. These studies have shown that an economic variable may react differently in response to downturns and recoveries in a business cycle. Inspiring from empirical results, this thesis investigates dynamics of Turkish key macroeconomic data, namely capacity utilization rate, growth of import and export volume indices, growth of gross domestic product, interest rate for cash loans in Turkish Liras and growth of industrial production index. Estimation results imply that capacity utilization rate and growth of industrial production index show M-TAR type nonlinear stationary behavior according to the unit root test proposed by Enders and Granger (1998).

Made available in DSpace on 2014-06-11T19:26:07Z (GMT). No. of bitstreams: 0 Previous issue date: 2008-04-18Bitstream added on 2014-06-13T20:33:49Z : No. of bitstreams: 1 galeazzi_pm_me_jabo.pdf: 271602 bytes, checksum: a23deef5dbc3886b20b583de5242ab23 (MD5) / Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) / Na expectativa de contribuir com os programas de melhoramento genético de bubalinos, o objetivo desse estudo foi verificar a influência dos efeitos ambientais e genéticos sobre a habilidade de permanência no rebanho de búfalas leiteiras da raça Murrah. Utilizaram-se informações de 1016 búfalas, participantes do programa de controle leiteiro mantido pelo Departamento de Zootecnia da UNESP/Jaboticabal(SP) desde 1987. A característica habilidade de permanência (HP) foi definida como a habilidade de permanecer no rebanho um (HP1), dois (HP2), três (HP3), quatro (HP4), cinco (HP5) e seis anos (HP6) após o primeiro parto. As mesmas características foram também consideradas como variável contínua, e denominadas Habilidade de Permanência em Dias até um (HPD1), dois (HPD2), três (HPD3), quatro (HPD4), cinco (HPD5) e seis anos (HPD6) após o primeiro parto. Os efeitos ambientais foram estudados por meio de análise de sobrevivência, ajustando-se os efeitos fixos de fazenda, ano e época de nascimento, classe de produção de leite na primeira lactação e idade ao primeiro parto. Fazenda, ano de nascimento e produção de leite na primeira lactação são efeitos que influenciam de forma significativa (p<0,0001) a permanência da fêmea no rebanho de 1 a 6 anos após o primeiro parto. Búfalas com maior idade ao primeiro parto têm maiores probabilidades de serem descartadas até um ano, sem efeito no descarte em idades mais avançadas. Búfalas com maior produção de leite têm menor probabilidade de descarte, permanecendo mais tempo no rebanho. Os efeitos genético-aditivo para as HP foram estimados por máxima verossimilhança restrita aproximada com modelo de limiar, enquanto que para as HPD foram estimados por máxima verossimilhança restrita... / In order of contributing to the buffalo genetic improvement programs , the objective of this study was to assess the influence of environmental and genetic effects on the stayability of the Murrah dairy buffaloes. Records from 1016 buffaloes, participants in the control dairy program maintained by the Departamento de Génetica e Melhoramento Animal of UNESP / Jaboticabal (SP) since 1987, were used. Stayability (HP) was defined as the ability to remain in the herd for one (HP1), two (HP2), three (HP3), four (HP4), five (HP5), and six years (HP6) after the first calving. The same traits were also considered continuous, and called as age of culling up to one (HPD1), two (HPD2), three (HPD3), four (HPD4), five (HPD5) and six years (HPD6) after the first calving. The environmental effects were studied by survival analysis. The model included the fixed effects of farm, year and season of birth, milk production class in the first lactation and age at first calving. The effects of farm, year of birth and milk production in the first lactation were significant (p <0.0001) for stayability from 1 to 6 years after the first calving.. Buffaloes cows with higher age at first calving are more likely to be culling a year after the first calve, without effect on the disposal at more advanced ages. Buffaloes with higher milk production are less likely to culling, staying longer in the herd. Additive genetic effects for HP were estimated by quasi-restricted maximum likelihood, fitting a threshold model, and the additive genetic effects for HPD were estimated by restricted maximum likelihood. The models included the additive genetic effects as random, ...(Complete abstract, click electronic access below)

La rémunération des grands dirigeants a augmenté de façon substantielle ces dernières décennies et a participé à l’accroissement des inégalités par les hauts revenus. Deux typologies pour les actionnaires de contrôle sont proposées pour analyser l’évolution et les pratiques de rémunérations des (P-)DG au sein des grands groupes cotés français. La première typologie se fonde sur l’identité de l’actionnaire et différencie des actionnaires actifs et engagés dans la stratégie de la firme, et des actionnaires passifs ou diversifiés. Parmi ces derniers se trouvent des actionnaires dont la stratégie est essentiellement financière. Les rémunérations sont plus élevées et plus sensibles aux performances de court-terme pour ces derniers. La seconde typologie se fonde sur deux critères : le degré (participation au capital) et l’ancienneté (années d’exercice du contrôle) du contrôle. En s’appuyant sur un modèle de données de panel à effet de seuil (modèle PTR), les politiques de rémunération des dirigeants permettent de différencier quatre régimes dans le degré, et deux dans l’ancienneté du contrôle. Cette typologie différencie les contrôles de type dispersé, influent, dominant et majoritaire, d’une part, et les effets des actionnaires de contrôle récents et de long-terme, d’autre part. L’évolution des rémunérations est enfin étudiée en se fondant sur ces deux typologies et sur une période de 12 années. Au-delà de facteurs liés au fonctionnement du « marché des dirigeants », les typologies mettent en évidence un phénomène de rattrapage et de contagion consécutif à la transparence des rémunérations depuis 2001, et une forte augmentation des bonus sous l’effet de l’augmentation des capitalisations boursières et des transformations de l’actionnariat qui l’ont accompagnée. / CEO compensation rose substantially in the last decades and contributed to the rise in high revenues. Two typologies for the controlling shareholders are suggested to study the evolution and the pay-setting process of CEO compensation in large French listed firms. The first typology is based on the identity of shareholders and differentiates active and involved ones to passive or diversified ones. While the former are involved in the industrial strategy of the firm, the latter comprise shareholders whose strategy is mainly financial. CEO pays are higher and more sensitive to short-term performance in this latter case. The second typology is defined according to two criteria, namely the degree and seniority of control. They respectively refer to the share of equity and the number of years of control. Based on a Panel Threshold Regression (PTR) model, CEO pay policies allows to differentiate four regimes in the degree of control, and two in the seniority of control, namely dispersed, influent, dominant, and majority control on the one hand, and new and long-term controlling shareholders, on the other hand. These two typologies are eventually used to study the evolution of CEO pay over a 12-year period in France. Beyond factors related to the so-called “market for CEOs”, the typologies reveal a catching up and contagion effect since the implementation of transparency since 2001, and a tremendous rise in bonuses that went hand in hand with the rise in market capitalizations and subsequent changes in ownership patterns.

Resumo: Na expectativa de contribuir com os programas de melhoramento genético de bubalinos, o objetivo desse estudo foi verificar a influência dos efeitos ambientais e genéticos sobre a habilidade de permanência no rebanho de búfalas leiteiras da raça Murrah. Utilizaram-se informações de 1016 búfalas, participantes do programa de controle leiteiro mantido pelo Departamento de Zootecnia da UNESP/Jaboticabal(SP) desde 1987. A característica habilidade de permanência (HP) foi definida como a habilidade de permanecer no rebanho um (HP1), dois (HP2), três (HP3), quatro (HP4), cinco (HP5) e seis anos (HP6) após o primeiro parto. As mesmas características foram também consideradas como variável contínua, e denominadas Habilidade de Permanência em Dias até um (HPD1), dois (HPD2), três (HPD3), quatro (HPD4), cinco (HPD5) e seis anos (HPD6) após o primeiro parto. Os efeitos ambientais foram estudados por meio de análise de sobrevivência, ajustando-se os efeitos fixos de fazenda, ano e época de nascimento, classe de produção de leite na primeira lactação e idade ao primeiro parto. Fazenda, ano de nascimento e produção de leite na primeira lactação são efeitos que influenciam de forma significativa (p<0,0001) a permanência da fêmea no rebanho de 1 a 6 anos após o primeiro parto. Búfalas com maior idade ao primeiro parto têm maiores probabilidades de serem descartadas até um ano, sem efeito no descarte em idades mais avançadas. Búfalas com maior produção de leite têm menor probabilidade de descarte, permanecendo mais tempo no rebanho. Os efeitos genético-aditivo para as HP foram estimados por máxima verossimilhança restrita aproximada com modelo de limiar, enquanto que para as HPD foram estimados por máxima verossimilhança restrita...(Resumo completo, clicar acesso eletrônico abaixo) / Abstract: In order of contributing to the buffalo genetic improvement programs , the objective of this study was to assess the influence of environmental and genetic effects on the stayability of the Murrah dairy buffaloes. Records from 1016 buffaloes, participants in the control dairy program maintained by the Departamento de Génetica e Melhoramento Animal of UNESP / Jaboticabal (SP) since 1987, were used. Stayability (HP) was defined as the ability to remain in the herd for one (HP1), two (HP2), three (HP3), four (HP4), five (HP5), and six years (HP6) after the first calving. The same traits were also considered continuous, and called as age of culling up to one (HPD1), two (HPD2), three (HPD3), four (HPD4), five (HPD5) and six years (HPD6) after the first calving. The environmental effects were studied by survival analysis. The model included the fixed effects of farm, year and season of birth, milk production class in the first lactation and age at first calving. The effects of farm, year of birth and milk production in the first lactation were significant (p <0.0001) for stayability from 1 to 6 years after the first calving.. Buffaloes cows with higher age at first calving are more likely to be culling a year after the first calve, without effect on the disposal at more advanced ages. Buffaloes with higher milk production are less likely to culling, staying longer in the herd. Additive genetic effects for HP were estimated by quasi-restricted maximum likelihood, fitting a threshold model, and the additive genetic effects for HPD were estimated by restricted maximum likelihood. The models included the additive genetic effects as random, ...(Complete abstract, click electronic access below) / Orientador: Humberto Tonhati / Coorientadora: Maria Eugênia Zerlotti Mercadante / Banca: Danísio Prado Munari / Banca: Lenira El Faro Zadra / Mestre

Phénomène social, culturel mais aussi économique, le tourisme s'est imposé au fil des décennies comme une source de revenus et de devises étrangères pour les pays hôtes au cours du 20e siècle. Cependant, l'impact du tourisme sur une économie à long terme et notamment une petite économie insulaire (PEI) reste encore ambigu.Cette thèse cherche à poser clairement la question de l’analyse des fondamentaux de la spécialisation touristique comme source soutenable de croissance pour une petite économie insulaire. Cette entreprise, réalisée en deux parties, nous conduit, d'une part, à redéfinir les principaux composants du tourisme en tant qu'activité sociale et économique, et à préciser comment le tourisme contribue à la croissance économique d’une PEI. D'autre part, la thèse lève le voile sur l'ambigüité de la relation entre tourisme et croissance. Elle fournit de nouveaux éléments d'explication d'un modèle de croissance par le tourisme plus vertueux et adapté au marché du tourisme mondial. Les résultats de l'introduction du tourisme dans un modèle théorique de croissance endogène à la Romer et d'une formalisation économétrique à seuil appliquée à un échantillon de PEI à l'aide de l'indice de pénétration touristique remettent en cause l'hypothèse de linéarité entre spécialisation touristique et croissance économique. Ils démontrent la fin d'un modèle de développement touristique intensif. Toutefois, cette situation n'est pas une fatalité. Notre analyse aboutit à la conclusion que le tourisme constitue une source soutenable de croissance pour une économie d'une PEI, lorsque l'intelligence et l'agilité forment les leitmotivs d'une stratégie touristique durable et multipartite. / Tourism has become a social, cultural and economic phenomenon. However, the impact of tourism on a long-term economy growth remains ambiguous, especially for small island economies (SIE).This thesis seeks to identify the fundamentals of tourism specialization as a sustainable source of growth for a SIE. On the one hand, it has led us to redefine the main components of tourism as a social and economic activity and to clarify how tourism contributes to the economic growth. On the other hand, this thesis reveals the ambiguity of the relationship between tourism and growth. It provides new explanations for a more virtuous tourism growth model and more adapted to the global tourism market.The results of the introduction of tourism into a Romer's model of endogenous growth and those of a threshold regression model applied to a sample of SEI call into question the hypothesis of linearity between tourist specialization and economic growth. The thesis demonstrates the end of a model of intensive tourist development. It concludes that tourism is a sustainable source of growth for an SIE, when intelligence and agility constitute the leitmotif of a sustainable, multi-stakeholder tourism strategy.

Estudos genético quantitativos e moleculares são fundamentais para o melhoramento animal e sua realização com a raça Nelore é de grande importância devido a ampla participação dessa no rebanho de corte nacional. A estimação constante dos parâmetros genéticos das características de produção é necessário para a adequada condução do processo de seleção dos animais. A melhoria de características relacionadas à carcaça bovina é essencial para a eficiência e sustentabilidade da atividade e a implementação de métodos de seleção animal baseados em informações moleculares pode revolucionar a produção zootécnica e deve ser profundamente estudado. Sendo assim, os objetivos do presente estudo foram estimar parâmetros genéticos e componentes de variância através de diferentes modelos matemáticos para um total de 14 características fenotípicas (o peso ao nascimento, peso a desmama, peso ao sobreano, ganho de peso entre a desmama e o sobreano ajustado para um intervalo de 345 dias, perímetro escrotal ao sobreano, altura de garupa ao sobreano, escores visuais avaliados ao sobreano de conformação, precocidade, musculosidade, comprimento de umbigo e ossatura, e ainda características de carcaça mensuradas por ultrassonografia realizada após 30 a 45 dias de confinamento como a área de olho de lombo, espessura de gordura subcutânea, espessura de gordura na picanha). Foram estimadas correlações entre todas estas características com as de carcaça mensuradas por ultrassonografia. Sob o enfoque molecular, desenvolveu-se um programa para imputação de genótipos faltantes e estudaram-se diferentes métodos de associação de marcadores moleculares do tipo mutação de base nitrogenada única (SNP) a características de produção incluídas no índice de seleção de um programa de melhoramento da raça Nelore, utilizando inferência bayesiana. Todas as características estudadas podem ser selecionadas esperando-se progresso genético na população. Os efeitos maternos foram importantes em algumas características onde normalmente estes efeitos não têm sido considerados atualmente. A quantidade de escores atribuídos a uma característica categórica assim como o número de observações fenotípicas resultam em diferenças nas estimativas quando avaliadas por modelos lineares ou de limiar. Não deverão ser obtidos resultados satisfatórios na melhoria das características de carcaça se a seleção for baseada nas tradicionais avaliações visuais utilizadas no momento. Os métodos utilizados na análise de associação dos marcadores podem originar diferentes resultados. Os marcadores que apresentaram efeitos altamente relevantes (P<0,01) geralmente apresentaram resultados semelhantes, independentemente do método utilizado. Certos marcadores podem ter efeitos positivos para algumas características componentes do índice de seleção e negativos para as demais. A análise em conjunto com todos os SNP\'s e todos os dados fenotípicos disponíveis é viável e parece ser a mais adequada. O método desenvolvido de imputação de genótipos faltantes a partir do parentesco de animais genotipados foi eficiente. / Quantitative and molecular genetic studies are very important for animal breeding and studies with Nellore cattle have great importance due to the large participation of that breed in the Brazilian beef cattle industry (around 80% of the herd). The constant estimation of genetic parameters for traits linked to production is necessary for properly perform selection of animals. The improvement of carcass traits is essential for efficiency and profitability of the activity. The implementation of methods for animal selection based on molecular information could revolutionize animal production and should be deeply studied. Thus, the objectives of this study were to estimate genetic parameters and variance components using different mathematical models for a total of 14 traits, such as birth weight, weaning weight, yearling weight, post-weaning weight gain between weaning and yearling adjusted for 345 days, yearling scrotal circumference, yearling hip height, yearling visual scores like conformation, finishing, muscularity, bone structure and navel length. Ultrasound measurements for carcass traits performed at feedlot (30 to 45 days, at approximate age of 20 months) such as rib-eye area, fat thickness, rump fat thickness were, also, evaluated. Estimate correlations between all these traits with the carcass traits measured by ultrasound were estimated. As concerned to molecular study, an algorithm for imputation of missing genotypes was developed and different methods to analyze molecular marker (single nucleotide polymorphism - SNP) association with traits components of the selection index of a breeding program that is applied to the population studied, using bayesian inference, were used. Genetic progress will be expected for selection of all the traits studied. Maternal effects were important in some traits in which those effects are not usually considered. The amount of scores assigned to a categorical trait and the number of observations could result in different estimates when evaluated by linear or threshold models. The selection for visual scores traditionally used in that population will not improve carcass traits. The methods used to analyze the association of markers may lead to different results. The SNP\'s with association effects of high relevance (P<0.01) generally express their effects regardless of the method used to analyze. Some markers may have positive effects for some traits of selection index, but negative for others. The joint analysis with all SNPs and with all available phenotypes is feasible and appears to be more appropriate. The algorithm developed for imputation of missing genotypes from pedigree information of genotyped animals was efficient.

Les caractéristiques des fluctuations des concentrations de polluants dans un environnement intérieur normalement occupé dépendent fortement de plusieurs paramètres, en particulier des occupants et de leurs activités et comportement, qui altèrent de manière considérable la nature statistique de leur variabilité temporelle. Ces fluctuations sont rarement disponibles et décrites dans la littérature. La mise à disposition de mesures en continu de la concentration en formaldéhyde et en particules (de 0,35 à 20 µm de diamètre) dans l’air d’un environnement de bureau et d’une maison expérimentale sur plusieurs mois avec un pas de temps fin (entre 1 min et 1 heure) a permis d’engager une réflexion sur la nature de ces fluctuations et leur prévisibilité. L’objectif de la thèse se décline en trois axes de recherche : (i) la caractérisation des fluctuations des concentrations des polluants cible ; (ii) la mise en évidence des sources de variabilité de ces fluctuations et (iii) la prévision des concentrations de ces polluants. Le premier axe concerne la détermination des caractéristiques communes partagées par les différents polluants. Le deuxième axe porte, à l'aide des approches par séparation aveugle des sources, sur l'estimation des déterminants des sources de variabilité. Le troisième axe est consacré à la prévision des fluctuations de concentration des polluants. L'analyse des séries temporelles pour ce type de données (hautes fréquences) doit prendre en compte l’échelle de temps sur laquelle évoluent plusieurs microstructures. Plusieurs outils ont été employés : l'analyse spectrale (dépendance à long terme par la mesure fractale et la statistique R/S), la mesure de l’oméga-prédictibilité, ainsi que la décomposition des séries en composantes latentes par STL (seasonal trend decomposition using Loess), SSA (singular spectrum analysis) et SBD (spectral band decomposition).L'identification des sources de variabilité de particules a été abordée par les méthodes de séparation aveugle des sources basées sur une factorisation matricielle en profils et contributions, sous contrainte statistique d'indépendance (ACI) ou de non-négativité (NNMF ou PMF). Les factorisations ont été appliquées à la matrice constituée des séries temporelles de différentes gammes de taille des particules. Certains profils ou contributions des sources ont pu été interprétés grâce aux variables exogènes "traceurs" de certaines sources (comme le CO2 indicateur de la présence de la source occupants).Concernant la prévision des concentrations de polluants, le choix des modèles a été basé sur la structure de la série temporelle, mise en évidence lors de l’étape de caractérisation. On constate de meilleures performances de prévisions sur la série ayant subi un prétraitement statistique de décomposition STL, SSA ou SBD. Plusieurs types de modèles stochastiques (linéaires ou non linéaires) ont été appliqués ensuite à ces composantes et le résultat final de la prévision est donné par leur combinaison. La caractéristique de non-linéarité qui apparaît sous forme de changements abrupts de concentration causée en grande partie par la manipulation des ouvrants et qui se greffe sur l'évolution régulière du système dynamique mérite un traitement spécial. Un nouveau type de modèles de prévision a été développé pour répondre aux exigences de la nature des données hautes fréquences présentant ce type de non linéarité. Ce modèle associe une étape de décomposition des séries en bandes spectrales (SBD) couplée avec une étape de modélisation par des modèles autorégressifs à seuil (TAR) ou par la dynamique du chaos : FFT-(TAR/Chaos).Les résultats montrent que le prétraitement par décomposition en bandes spectrales ou STL améliore sensiblement la prévision des concentrations de formaldéhyde et des particules fines jusqu’à un horizon de 10 heures pour le formaldéhyde (pas de temps 1 minute) et de 1 à 4 jours pour les particules selon leur taille (pas de temps horaire) / The time fluctuation features of particulate concentrations in a real occupied indoor environment are strongly dependent of several parameters and in particular the occupation and occupants’ activities and behaviors. These parameters considerably alter the statistical variability of the time series dynamics. These fluctuations are rarely available and described in the literature. These types of fluctuations are rarely available and described in the literature. The availability of continuous measurements of concentrations of formaldehyde and particulate matter (from 0.35 to 20 µm of diameter) in an indoor environment (office and a test house) during several months with a fine time resolution (from 1 min to 1 hour) permitted to initiate a process of reflection on the nature of these fluctuations and their predictability. The aim of the thesis follows three main axis of research: (i) characterizing indoor environment pollutant concentrations variability; (ii) revealing the sources of variability of these fluctuations and (iii) forecasting the pollutant concentrations in a real indoor environment. The first axis concerns the determination of the common features shared by the different pollutants. The second axis focuses on the sources variability estimating using a Blind Source Separation (BSS) approach. Lastly, the third axis focuses on the forecasting of pollutant concentrations. The time series analysis for this type of data (high frequency) should take into account the time scale on which microstructures evolve. Several tools were employed, such as the spectral analysis (long-range dependency by fractal dimension measures and R/S statistic), the omega-predictability, as well as the time series decomposition into latent components by STL (Seasonal Trend Decomposition using Loess), SSA (Singular Spectrum Analysis) and SBD (Spectral Band Decomposition).The identification of the sources of particles concentrations is developed using BSS based methods which are based on a matrix factorization as profiles and contributions under a statistical independence constraint (ICA) or a non-negativity constraint (NNMF or PMF). The factorizations were applied to the matrix of the time series of different particle size bins. Some profiles or source contributions could be interpreted using exogenous variables as fingerprints of some sources (such as CO2 concentrations used as indicator of the “occupation”).Concerning the pollutant concentrations forecasting, the model selection was made in agreement with the time series structures, highlighted in the characterization stage. One can notice better performances forecasts when using the series having been preprocessed by decomposition: STL, SSA or decomposition in spectral bands (based Transform Fourier), SBD. Several types of stochastic models (linear or nonlinear) were then applied to these components and the final forecast result is given by their combination. A special type of nonlinearities involving a special treatment is the abrupt concentration changes in time series concentrations due mainly to windows manipulation and graft on the regular evolution of the dynamic system. A new type of forecast models has been developed in adequacy with the requirements of the nature of high-frequency data. This model combine spectral band decomposition step (SBD) coupled with a modeling stage based on autoregressive switching threshold model (TAR) or chaos dynamic: FFT- (TAR / Chaos).The results show that the pretreatment by spectral band decomposition or STL improves significantly formaldehyde and fine particles concentrations forecast on 10-hour horizon for formaldehyde (sampled every minute) and on a horizon of 1 to 4 days for the particles (sampled every hour), depending on their size bins

本文探討退休金發展程度與投資策略對股票市場發展的影響，並同時採用「縱橫門檻迴歸模型」（panel threshold model, PTM）及結合縱橫門檻模型與穩健迴歸的「穩健縱橫門檻迴歸模型」（robust panel threshold model, ROPTM）來研究此一議題。我們用退休基金投資證券市場的金額佔總額的比例為分類標準，將樣本分為高投資比例與低投資比例兩部分。對部分OECD國家及台灣的panel data分析後之結果顯示：在股票市場方面，若基金採高投資比例之投資策略，則退休金發展或投資股市比例越高，越能促進股市發展；採低投資比例策略的基金，對股市發展的影響並不顯著。 / This paper analyzes the impact of pension fund investment on securities markets using a panel threshold model (PTM) and a robust panel threshold model (ROPTM) which combines a panel threshold model with a robust regression model. We use panel data for some OECD countries and Taiwan to test the validity of our propositions. The data is divided into low and high investment regions based on the value of securities as a percentage of total financial assets of the pension fund. Our results are the following. In the high stock investment region, pension funds have a positive impact on stock markets. Whereas, in the low stock investment region, the positive impact seems to disappear.