Tesi sul tema "Extremal dependence modeling"

Offshore structures, such as oil platforms and vessels, must be built such that they can withstand extreme environmental conditions (e.g., high waves and strong winds) that may occur during their lifetime. This means that it is essential to quantify probabilities of the occurrence of such extreme events. However, a difficulty arises in that there are very limited data available at these levels. The statistical field of extreme value theory provides asymptotically motivated models for extreme events, hence allowing extrapolation to very rare events. In addition to the risk to a single site, we are also interested in the joint risk of multiple offshore platforms being affected by the same extreme event. In order to understand joint extremal behaviour for two or more locations, the spatial dependence between the different locations must be considered. Extremal dependence between two locations can be of two types: asymptotic independence (AI) when the extremes at the two sites are unlikely to occur together, and asymptotic dependence (AD) when it is possible for both sites to be affected simultaneously. For finite samples it is often difficult to determine which type of dependence the data are more consistent with. In a large ocean basin it is reasonable to expect both of these features to be present, with some close by locations AD, with the dependence decreasing with distance, and some far apart locations AI. In this thesis we develop new diagnostic tools for distinguishing between AD and AI and illustrate these on North Sea wave height data. We also investigate how extremal dependence changes with direction and find evidence for spatial anisotropy in our data set. The most widely used spatial models assume asymptotic dependence or perfect independence between sites, which is often unrealistic in practice. Models that attempt to capture both AD and AI exist, but they are difficult to implement in practice due to their complexity and they are restricted in the forms of AD and AI they can model. In this thesis we introduce a family of bivariate distributions that exhibits all the required features of short, medium and long range extremal dependence required for pairwise dependence modelling in spatial applications.

In financial risk management it is essential to be able to model dependence in markets and portfolios in an accurate and efficient way. A high positive dependence between assets in a portfolio can be devastating, especially in times of crises, since losses will most likely occur at the same time in all assets for such a portfolio. The dependence is therefore directly linked to the risk of the portfolio. The risk can be estimated by several different risk measures, for example Value-at-Risk and Expected shortfall. This paper studies some different ways to measure risk and model dependence, both in a theoretical and empirical way. The main focus is on copulas, which is a way to model and construct complex dependencies. Copulas are a useful tool since it allows the user to separately specify the marginal distributions and then link them together with the copula. However, copulas can be quite complex to understand and it is not trivial to know which copula to use. An implemented copula model might give the user a "black-box" feeling and a severe model risk if the user trusts the model too much and is unaware of what is going. Another model would be to use the linear correlation which is also a way to measure dependence. This is an easier model and as such it is believed to be easier for all users to understand. However, linear correlation is only easy to understand in the case of elliptical distributions, and when we move away from this assumption (which is usually the case in financial data), some clear drawbacks and pitfalls become present. A third model, called historical simulation, uses the historical returns of the portfolio and estimate the risk on this data without making any parametric assumptions about the dependence. The dependence is assumed to be incorporated in the historical evolvement of the portfolio. This model is very easy and very popular, but it is more limited than the previous two models to the assumption that history will repeat itself and needs much more historical observations to yield good results. Here we face the risk that the market dynamics has changed when looking too far back in history. In this paper some different copula models are implemented and compared to the historical simulation approach by estimating risk with Value-at-Risk and Expected shortfall. The parameters of the copulas are also investigated under calm and stressed market periods. This information about the parameters is useful when performing stress tests. The empirical study indicates that it is difficult to distinguish the parameters between the stressed and calm market period. The overall conclusion is; which model to use depends on our beliefs about the future distribution. If we believe that the distribution is elliptical then a correlation model is good, if it is believed to have a complex dependence then the user should turn to a copula model, and if we can assume that history will repeat itself then historical simulation is advantageous.

Market risk modelling is one of the most dynamic domains in finance. Risk is the uncertainty that affects the values of assets in the system in an unknown fashion causing fluctuations in their values and in investment outcomes. Market risk is defined as the losses due to fluctuations in the prices of financial assets which are caused by changing market conditions. Market risk modelling comprises tools and techniques which quantify the risk associated with financial instruments. Risk quantification is necessary to devise strategies such as hedging or diversification against the risk, to avoid severe losses. With the recent financial market events like the Global Financial Crisis, there is a need to evaluate the traditional risk return relationships presented in Asset Pricing models and more sophisticated risk modelling tools like Value at Risk (VaR). Along with Asset Pricing and VaR modelling another important risk issue between financial assets is the asymptotic tail dependence, which plays a vital role in accurate risk measurement in portfolio selection and hedging amongst other considerations. The usual measure of dependence, the Pearson Correlation coefficient works on the assumption of normality in the data distribution and hence is unable to capture the tail dependence between financial assets which is an important characteristic for tail risk modelling. The research presented in this dissertation models the risk quantification techniques of Asset Pricing, VaR modelling and Tail dependence, with the more sophisticated statistical tools of Quantile Regression and Extreme Value Theory (EVT), which are particularly useful in modelling the tail behaviour of the distributions. The research targets four broad objectives to evaluate extreme risk and dependence measures in the Australian stock market which are realised with the robust techniques of Quantile Regression and EVT. The thesis comprises six chapters with chapter-1 introducing the thesis presenting the driving motivations for the research and the four major objectives (which are detailed in individual chapters following chapter-1) along with the contribution of the research and finally chapter-6 presenting the conclusion. The structure of rest of the thesis is also outlined in chapter-1.

Dans un grand éventail d'applications allant des sciences du climat à la finance, des événements extrêmes avec une probabilité loin d'être négligeable peuvent se produire, entraînant des conséquences désastreuses. Les extrêmes d'évènements climatiques tels que le vent, la température et les précipitations peuvent profondément affecter les êtres humains et les écosystèmes, entraînant des événements tels que des inondations, des glissements de terrain ou des vagues de chaleur. Lorsque l'emphase est mise sur l'étude de variables mesurées dans le temps sur un grand nombre de stations ayant une localisation spécifique, comme les variables mentionnées précédemment, le partitionnement de variables devient essentiel pour résumer et visualiser des tendances spatiales, ce qui est crucial dans l'étude des événements extrêmes. Cette thèse explore plusieurs modèles et méthodes pour partitionner les variables d'un processus stationnaire multivarié, en se concentrant sur les dépendances extrémales.Le chapitre 1 présente les concepts de modélisation de la dépendance via les copules, fondamentales pour la dépendance extrême. La notion de variation régulière est introduite, essentielle pour l'étude des extrêmes, et les processus faiblement dépendants sont abordés. Le partitionnement est discuté à travers les paradigmes de séparation-proximité et de partitionnement basé sur un modèle. Nous abordons aussi l'analyse non-asymptotique pour évaluer nos méthodes dans des dimensions fixes.Le chapitre 2 est à propos de la dépendance entre valeurs maximales est cruciale pour l'analyse des risques. Utilisant la fonction de copule de valeur extrême et le madogramme, ce chapitre se concentre sur l'estimation non paramétrique avec des données manquantes. Un théorème central limite fonctionnel est établi, démontrant la convergence du madogramme vers un processus Gaussien tendu. Des formules pour la variance asymptotique sont présentées, illustrées par une étude numérique.Le chapitre 3 propose les modèles asymptotiquement indépendants par blocs (AI-blocs) pour le partitionnement de variables, définissant des clusters basés sur l'indépendance des maxima. Un algorithme est introduit pour récupérer les clusters sans spécifier leur nombre à l'avance. L'efficacité théorique de l'algorithme est démontrée, et une méthode de sélection de paramètre basée sur les données est proposée. La méthode est appliquée à des données de neurosciences et environnementales, démontrant son potentiel.Le chapitre 4 adapte des techniques de partitionnement pour analyser des événements extrêmes composites sur des données climatiques européennes. Les sous-régions présentant une dépendance des extrêmes de précipitations et de vitesse du vent sont identifiées en utilisant des données ERA5 de 1979 à 2022. Les clusters obtenus sont spatialement concentrés, offrant une compréhension approfondie de la distribution régionale des extrêmes. Les méthodes proposées réduisent efficacement la taille des données tout en extrayant des informations cruciales sur les événements extrêmes.Le chapitre 5 propose une nouvelle méthode d'estimation pour les matrices dans un modèle linéaire à facteurs latents, où chaque composante d'un vecteur aléatoire est exprimée par une équation linéaire avec des facteurs et du bruit. Contrairement aux approches classiques basées sur la normalité conjointe, nous supposons que les facteurs sont distribués selon des distributions de Fréchet standards, ce qui permet une meilleure description de la dépendance extrémale. Une méthode d'estimation est proposée garantissant une solution unique sous certaines conditions. Une borne supérieure adaptative pour l'estimateur est fournie, adaptable à la dimension et au nombre de facteurs
In a wide range of applications, from climate science to finance, extreme events with a non-negligible probability can occur, leading to disastrous consequences. Extremes in climatic events such as wind, temperature, and precipitation can profoundly impact humans and ecosystems, resulting in events like floods, landslides, or heatwaves. When the focus is on studying variables measured over time at numerous specific locations, such as the previously mentioned variables, partitioning these variables becomes essential to summarize and visualize spatial trends, which is crucial in the study of extreme events. This thesis explores several models and methods for partitioning the variables of a multivariate stationary process, focusing on extreme dependencies.Chapter 1 introduces the concepts of modeling dependence through copulas, which are fundamental for extreme dependence. The notion of regular variation, essential for studying extremes, is introduced, and weakly dependent processes are discussed. Partitioning is examined through the paradigms of separation-proximity and model-based clustering. Non-asymptotic analysis is also addressed to evaluate our methods in fixed dimensions.Chapter 2 study the dependence between maximum values is crucial for risk analysis. Using the extreme value copula function and the madogram, this chapter focuses on non-parametric estimation with missing data. A functional central limit theorem is established, demonstrating the convergence of the madogram to a tight Gaussian process. Formulas for asymptotic variance are presented, illustrated by a numerical study.Chapter 3 proposes asymptotically independent block (AI-block) models for partitioning variables, defining clusters based on the independence of maxima. An algorithm is introduced to recover clusters without specifying their number in advance. Theoretical efficiency of the algorithm is demonstrated, and a data-driven parameter selection method is proposed. The method is applied to neuroscience and environmental data, showcasing its potential.Chapter 4 adapts partitioning techniques to analyze composite extreme events in European climate data. Sub-regions with dependencies in extreme precipitation and wind speed are identified using ERA5 data from 1979 to 2022. The obtained clusters are spatially concentrated, offering a deep understanding of the regional distribution of extremes. The proposed methods efficiently reduce data size while extracting critical information on extreme events.Chapter 5 proposes a new estimation method for matrices in a latent factor linear model, where each component of a random vector is expressed by a linear equation with factors and noise. Unlike classical approaches based on joint normality, we assume factors are distributed according to standard Fréchet distributions, allowing a better description of extreme dependence. An estimation method is proposed, ensuring a unique solution under certain conditions. An adaptive upper bound for the estimator is provided, adaptable to dimension and the number of factors

Dans cette thèse, nous abordons l'estimation non paramétrique de la fonction de dépendance des distributions multivariées à valeurs extrêmes. Dans une première partie, on adopte l’hypothèse classique stipulant que les variables aléatoires sont indépendantes et identiquement distribuées (i.i.d). Plusieurs estimateurs non paramétriques sont comparés pour une fonction de dépendance trivariée de type logistique dans deux différents cas. Dans le premier cas, on suppose que les fonctions marginales sont des distributions généralisées à valeurs extrêmes. La distribution marginale est remplacée par la fonction de répartition empirique dans le deuxième cas. Les résultats des simulations Monte Carlo montrent que l'estimateur Gudendorf-Segers (Gudendorf et Segers, 2011) est plus efficient que les autres estimateurs pour différentes tailles de l’échantillon. Dans une deuxième partie, on ignore l’hypothèse i.i.d vue qu’elle n'est pas vérifiée dans l'analyse des séries temporelles. Dans le cadre univarié, on examine le comportement extrêmal d'un modèle autorégressif Gaussien stationnaire. Dans le cadre multivarié, on développe un nouveau théorème qui porte sur la convergence asymptotique de l'estimateur de Pickands vers la fonction de dépendance théorique. Ce fondement théorique est vérifié empiriquement dans les cas d’indépendance et de dépendance asymptotique. Dans la dernière partie de la thèse, l'estimateur Gudendorf-Segers est utilisé pour modéliser la structure de dépendance des concentrations extrêmes d’ozone observées dans les stations qui enregistrent des dépassements de la valeur guide et limite de la norme Tunisienne de la qualité d'air NT.106.04
In this thesis, we investigate the nonparametric estimation of the dependence function for multivariate extreme value distributions. Firstly, we assume independent and identically distributed random variables (i.i.d). Several nonparametric estimators are compared for a trivariate dependence function of logistic type in two different cases. In a first analysis, we suppose that marginal functions are generalized extreme value distributions. In a second investigation, we substitute the marginal function by the empirical distribution function. Monte Carlo simulations show that the Gudendorf-Segers (Gudendorf and Segers, 2011) estimator outperforms the other estimators for different sample sizes. Secondly, we drop the i.i.d assumption as it’s not verified in time series analysis. Considering the univariate framework, we examine the extremal behavior of a stationary Gaussian autoregressive process. In the multivariate setting, we prove the asymptotic consistency of the Pickands dependence function estimator. This theoretical finding is confirmed by empirical investigations in the asymptotic independence case as well as the asymptotic dependence case. Finally, the Gudendorf-Segers estimator is used to model the dependence structure of extreme ozone concentrations in locations that record several exceedances for both guideline and limit values of the Tunisian air quality standard NT.106.04

The thesis aims to investigate some of the approaches to modelling portfolio returns with heavy-tailed risk factors. It first elaborates on the univariate time series models, and compares the benchmark model (GARCH with Student t innovations or its GJR extension) predictive performance with its two competitors, the EVT-GARCH model and the Markov-Switching Multifractal (MSM) model. The motivation of EVT extension of GARCH specification is to use a more proper distribution of the innovations, based on the empirical distribution function. The MSM is one of the best performing models in the multifractal literature, a markov-switching model which is unique by its parsimonious specification and variability. The performance of these models is assessed with Mincer-Zarnowitz regressions as well as by comparison of quality of VaR and expected shortfall predictions, and the empirical analysis shows that for the risk management purposes the EVT-GARCH dominates the benchmark as well as the MSM. The second part addresses the dependence structure modelling, using the Gauss and t-copula to model the portfolio returns and compares the result with the classic variance-covariance approach, concluding that copulas offer a more realistic estimates of future extreme quantiles.

Le comportement extrême joint entre variables aléatoires revêt un intérêt particulier dans de nombreuses applications des sciences de l’environnement, de la finance, de l’assurance ou encore de la gestion du risque. Par exemple, ce comportement joue un rôle central dans l’évaluation des risques de catastrophes naturelles. Une erreur de spécification de la dépendance entre des variables aléatoires peut engendrer une sous-estimation dangereuse du risque, en particulier au niveau extrême. Le premier objectif de cette thèse est de développer des techniques d’inférence pour les copules Archimax. Ces modèles de dépendance peuvent capturer tout type de dépendance asymptotique entre les extrêmes et, de manière simultanée, modéliser les risques joints au niveau moyen. Une copule Archimax est caractérisée par ses deux paramètres fonctionnels, la fonction de dépendance caudale stable et le générateur Archimédien qui agit comme une distorsion affectant le régime de dépendance extrême. Des conditions sont dérivées afin que le générateur et la fonction caudale soient identifiables, de sorte qu’une approche d’inférence semi-paramétrique puisse être développée. Deux estimateurs non paramétriques de la fonction caudale et un estimateur du générateur basé sur les moments, supposant que ce dernier appartient à une famille paramétrique, sont avancés. Le comportement asymptotique de ces estimateurs est ensuite établi sous des hypothèses de régularité non restrictives et la performance en échantillon fini est évaluée par le biais d’une étude de simulation. Une construction hiérarchique (ou en “clusters”) qui généralise les copules Archimax est proposée afin d’apporter davantage de flexibilité, la rendant plus adaptée aux applications pratiques. Le comportement extrême de ce nouveau modèle de dépendance est ensuite étudié, ce qui engendre un nouvelle manière de construire des fonctions de dépendance caudale stable. La copule Archimax est ensuite utilisée pour analyser les maxima mensuels de précipitations, observées à trois stations météorologiques en Bretagne. Le modèle semble très bien ajusté aux données, aussi bien aux précipitations faibles qu’aux précipitationsfortes. L’estimateur non paramétrique de la fonction caudale révèle une dépendance extrême asymétrique entre les stations, ce qui reflète le déplacement des orages dans la région. Une application du modèle Archimax hiérarchique à un jeu de données de précipitations contenant 155 stations est ensuite présentée, dans laquelle des groupes de stations asymptotiquement dépendantes sont déterminés via un algorithme de “clustering” spécifiquement adapté au modèle. Enfin, de possibles méthodes pour modéliser la dépendance inter-cluster sont évoquées
In various applications in environmental sciences, finance, insurance or risk management, joint extremal behavior between random variables is of particular interest. For example, this plays a central role in assessing risks of natural disasters. Misspecification of the dependence between random variables can lead to substantial underestimation of risk, especially at extreme levels. This thesis develops inference techniques for Archimax copulas. These copula models can account for any type of asymptotic dependence between extremes and at the same time capture joint risks at medium levels. An Archimax copula is characterized by two functional parameters, the stable tail dependence function (stdf), and the Archimedean generator which acts as a distortion of the extreme-value dependence model. Conditions under which the generator and the stdf are identifiable are derived so that a semiparametric approach for inference can be developed. Two nonparametric estimators of the stdf and a moment-based estimator of the generator, which assumes that the latter belongs to a parametric family, are proposed. The asymptotic behavior of the estimators is then established under broad regularity conditions; performance in small samples is assessed through a comprehensive simulation study. In the second part of the thesis, Archimax copulas are generalized to a clustered constructions in order to bring in more flexibility, which is needed in practical applications. The extremal behavior of this new dependence model is derived herein. Finally, the methodology proposed herein is illustrated on precipitation data. First, a trivariate Archimax copula is used to analyze monthly rainfall maxima at three stations in French Brittany. The model is seen to fit the data very well, both in the lower and in the upper tail. The nonparametric estimator of the stdf reveals asymmetric extremal dependence between the stations, which reflects heavy precipitation patterns in the area. An application of the clustered Archimax model to a precipitation dataset containing 155 stations is then presented, where groups of asymptotically dependent stations are determined via a specifically tailored clustering algorithm. Finally, possible ways to model inter cluster dependence are discussed

L'entrée en application depuis le 1er Janvier 2016 de la réforme réglementaire européenne du secteur des assurances Solvabilité 2 est un événement historique qui va changer radicalement les pratiques en matière de gestion des risques. Elle repose sur une prise en compte importante du profil et de la vision du risque, via la possibilité d'utiliser des modèles internes pour calculer les capitaux de solvabilité et l'approche ORSA (Own Risk and Solvency Assessment) pour la gestion interne du risque. La modélisation mathématique est ainsi un outil indispensable pour réussir un exercice réglementaire. La théorie du risque doit être en mesure d'accompagner ce développement en proposant des réponses à des problématiques pratiques, liées notamment à la modélisation des dépendances et aux choix des mesures de risques. Dans ce contexte, cette thèse présente une contribution à l'amélioration de la gestion des risques actuariels. En quatre chapitres nous présentons des mesures multivariées de risque et leurs applications à l'allocation du capital de solvabilité. La première partie de cette thèse est consacrée à l'introduction et l'étude d'une nouvelle famille de mesures multivariées élicitables de risque qu'on appellera des expectiles multivariés. Son premier chapitre présente ces mesures et explique les différentes approches utilisées pour les construire. Les expectiles multivariés vérifient un ensemble de propriétés de cohérence que nous abordons aussi dans ce chapitre avant de proposer un outil d'approximation stochastique de ces mesures de risque. Les performances de cette méthode étant insuffisantes au voisinage des niveaux asymptotiques des seuils des expectiles, l'analyse théorique du comportement asymptotique est nécessaire, et fera le sujet du deuxième chapitre de cette partie. L'analyse asymptotique est effectuée dans un environnement à variations régulières multivariées, elle permet d'obtenir des résultats dans le cas des queues marginales équivalentes. Nous présentons aussi dans le deuxième chapitre le comportement asymptotique des expectiles multivariés sous les hypothèses précédentes en présence d'une dépendance parfaite, ou d'une indépendance asymptotique, et nous proposons à l'aide des statistiques des valeurs extrêmes des estimateurs de l'expectile asymptotique dans ces cas. La deuxième partie de la thèse est focalisée sur la problématique de l'allocation du capital de solvabilité en assurance. Elle est composée de deux chapitres sous forme d'articles publiés. Le premier présente une axiomatisation de la cohérence d'une méthode d'allocation du capital dans le cadre le plus général possible, puis étudie les propriétés de cohérence d'une approche d'allocation basée sur la minimisation d'indicateurs multivariés de risque. Le deuxième article est une analyse probabiliste du comportement de cette dernière approche d'allocation en fonction de la nature des distributions marginales des risques et de la structure de la dépendance. Le comportement asymptotique de l'allocation est aussi étudié et l'impact de la dépendance est illustré par différents modèles marginaux et différentes copules. La présence de la dépendance entre les différents risques supportés par les compagnies d'assurance fait de l'approche multivariée une réponse plus appropriée aux différentes problématiques de la gestion des risques. Cette thèse est fondée sur une vision multidimensionnelle du risque et propose des mesures de nature multivariée qui peuvent être appliquées pour différentes problématiques actuarielles de cette nature
The entry into force since January 1st, 2016 of Solvency 2, the European regulatory reform of insurance industry, is a historic event that will radically change the practices in risk management. It is based on taking into account the own risk profile and the internal view of risk through the ability to use internal models for calculating solvency capital requirement and ORSA (Own Risk and Solvency Assessment) approach for internal risk management. It makes the mathematical modeling an essential tool for a successful regulatory exercise. The risk theory must allow to support this development by providing answers to practical problems, especially those related to the dependence modeling and the choice of risk measures. In the same context, this thesis presents a contribution to improving the management of insurance risks. In four chapters we present multivariate risk measures and their application to the allocation of solvency capital. The first part of this thesis is devoted to the introduction and study of a new family of multivariate elicitable risk measures that we will call multivariate expectiles. The first chapter presents these measures and explains the different construction approaches. The multivariate expectiles verify a set of coherence properties that we also discuss in this chapter before proposing a stochastic approximation tool of these risk measures. The performance of this method is insufficient in the asymptotic levels of the expectiles thresholds. That makes the theoretical analysis of the asymptotic behavior necessary. The asymptotic behavior of multivariate expectiles is then the subject of the second chapter of this part. It is studied in a multivariate regular variations framework, and some results are given in the case of equivalent marginal tails. We also study in the second chapter of the first part the asymptotic behavior of multivariate expectiles under previous assumptions in the presence of a perfect dependence, or in the case of asymptotic independence. Finally, we propose using extreme values statistics some estimators of the asymptotic expectile in these cases. The second part of the thesis is focused on the issue of solvency capital allocation in insurance. It is divided into two chapters; each chapter consists of a published paper. The first one presents an axiomatic characterization of the coherence of a capital allocation method in a general framework. Then it studies the coherence properties of an allocation approach based on the minimization of some multivariate risk indicators. The second paper is a probabilistic analysis of the behavior of this capital allocation method based on the nature of the marginal distributions of risks and the dependence structure. The asymptotic behavior of the optimal allocation is also studied and the impact of dependence is illustrated using some selected models and copulas. Faced to the significant presence of dependence between the various risks taken by insurance companies, a multivariate approach seems more appropriate to build responses to the various issues of risk management. This thesis is based on a multidimensional vision of risk and proposes some multivariate risk measures that can be applied to several actuarial issues of a multivariate nature

Analysis of extremes in a complex interacting hydroclimatic system is challenging due to rarity of events, small sample size, poor state of data or missing data, lack of statistical tools for analyzing observed changes and poor understanding of interactions of extremes. This thesis addresses the challenge of lack of statistical tools specifically suited for modeling hydroclimatic extremes. Significant changes are observed in the intensity, frequency, duration, timing and spatial extent of hydroclimatic extremes. Understanding the nature of these changing characteristics is a major challenge for hydroclimatic research community. Recent high impact extremes provide strong evidences for the interconnections of extremes. The boundaries of system are widening with the changes leading to a greater need to consider the dependencies between interacting processes for reliable risk estimates. A contribution of this thesis is to identify statistical procedures for modeling the extremal dependence structure in order to facilitate accurate probabilistic characterization of hydro-meteorological processes. Extreme rainfall is the most common cause of flooding and likelihood of such events is found to increase in recent studies. Large scale natural variability, human induced global warming and local atmospheric warming are important drivers of extreme rainfall. A methodology to investigate the association of daily rainfall extremes with plausible physical drivers is presented in this thesis. Non-stationary extreme value models are used to investigate the association and Copula theory is used to capture the dependence structure of extreme rainfall with the most significant physical driver. A combination of multiple processes can lead to devastating consequences, making the recovery of the system more difficult. Common statistical modeling practices or conceptual frameworks cannot capture the interrelationships of multiple extreme events which mutually enhance each-other. A statistical procedure to investigate the changes in the characteristics of concurrent meteorological droughts and heatwaves is presented in the thesis. Changes in the frequency and spatial extent of concurrent extremes are quantified over India to identify the hotspots which need immediate attention. The complex nature of concurrent extremes requires a new perspective for robust risk assessment. This thesis identifies parametric multivariate extreme value models as a suitable tool to model the dependence structure of concurrent extremes and disentangle their complex interactions. Extremal dependence structure of rainfall deficits, soil moisture deficits and high temperatures is explicitly described through angular densities on the two-dimensional simplex. These models can provide a powerful new perspective for appropriate statistical analysis of dependent hydroclimatic extremes in higher dimensions. Clustering of extreme rainfall in short period of time is responsible for huge economic and environmental losses. Extreme value models have been widely used to model the magnitudes of extreme events; however, little attention is paid to the duration of the extreme events due to challenges in modeling the dependence within the clusters of exceedances. This thesis presents a hierarchical Bayesian model to capture the temporal dependence structure of extreme rainfall spells. Specifically, this model addresses the risk of a flooding situation which arises due to heavy rainfall for a few consecutive days. The work presented in this thesis emphasizes the necessity of capturing the dependence structure of extremes to improve the understanding, modeling and prediction of hydroclimatic extremes.

This dissertation contains three essays concerning two broad areas, namely, optimal capital structure and risky assets modeling. In the first paper, we study corporate debt values, capital structure, and the term structure of interest rates in a unified framework. We employ numerical techniques to compute the firm's optimal capital structure and the value of its long-term risky debt with call option embedded and yield spreads when the value of the firm's unleveraged assets and the instantaneous default-free interest rate are risk factors. Debt and leveraged firm value are thus explicitly linked to properties of the firm's unleveraged assets, the term structure of default-free interest rates, taxes, bankruptcy costs, payout rates, and bond covenants. The results clarify the relationship between a firm's capital structure and movements in the term structure and other important aspects of the capital structure decision. In the second chapter, we propose a dynamic copula modeling framework that allows copula association parameters to change with time and macroeconomic variables. We find empirical evidence that nominal interest rate and price index for traded goods differentials between two countries have significant impact on the co-movement of foreign exchange rates. Our Pearson-type goodness-of-fit test has the power to reject constant and time-varying copula modeling approaches at the 95% confidence level. In the third chapter, a new method for solving sample size problem in probabilistic risk assessment has been developed. We propose the use of Bayesian power prior distributions to improve extreme value theory and provide reliable estimates of Value-at-Risk (VaR) and expected shortfall. The Bayesian Monte Carlo Markov chain computational scheme with power prior distributions allows us to properly incorporate historical data and borrow strength and information from related sources to current study.

abstract: To extend the lifetime of complementary metal-oxide-semiconductors (CMOS), emerging process techniques are being proposed to conquer the manufacturing difficulties. New structures and materials are proposed with superior electrical properties to traditional CMOS, such as strain technology and feedback field-effect transistor (FB-FET). To continue the design success and make an impact on leading products, advanced circuit design exploration must begin concurrently with early silicon development. Therefore, an accurate and scalable model is desired to correctly capture those effects and flexible to extend to alternative process choices. For example, strain technology has been successfully integrated into CMOS fabrication to improve transistor performance but the stress is non-uniformly distributed in the channel, leading to systematic performance variations. In this dissertation, a new layout-dependent stress model is proposed as a function of layout, temperature, and other device parameters. Furthermore, a method of layout decomposition is developed to partition the layout into a set of simple patterns for model extraction. These solutions significantly reduce the complexity in stress modeling and simulation. On the other hand, semiconductor devices with self-feedback mechanisms are emerging as promising alternatives to CMOS. Fe-FET was proposed to improve the switching by integrating a ferroelectric material as gate insulator in a MOSFET structure. Under particular circumstances, ferroelectric capacitance is effectively negative, due to the negative slope of its polarization-electrical field curve. This property makes the ferroelectric layer a voltage amplifier to boost surface potential, achieving fast transition. A new threshold voltage model for Fe-FET is developed, and is further revealed that the impact of random dopant fluctuation (RDF) can be suppressed. Furthermore, through silicon via (TSV), a key technology that enables the 3D integration of chips, is studied. TSV structure is usually a cylindrical metal-oxide-semiconductors (MOS) capacitor. A piecewise capacitance model is proposed for 3D interconnect simulation. Due to the mismatch in coefficients of thermal expansion (CTE) among materials, thermal stress is observed in TSV process and impacts neighboring devices. The stress impact is investigated to support the interaction between silicon process and IC design at the early stage.
Dissertation/Thesis
Ph.D. Electrical Engineering 2011

碩士
國立高雄應用科技大學
金融資訊研究所
95
Value at Risk is a widespread tool of risk management recently. It is a value that measures the worst loss of asset under the particular confidence level and possessed of period. Moreover, it is a quantile describing the tail of distribution of financial return series in statistics. In empirical literature, most of financial data have some properties such as fat tails and volatility clustering. Thus, estimating Value at Risk by conventional method may underestimate the quantile as a result of fat tails. We estimate Value at Risk in stock market by using extreme value theory combine with time series models and thereby compared the performance of the conditional Value at Risk with unconditional Value at Risk. In addition, we investigate optimal threshold level among past experience, method argued by hall and method proposed by Danielsson. Then we experiment backtesting on Value at Risk estimator and evaluate efficiency of estimator by LR statistic. We backtest mentioned above on eleven stock indexes：Dow Jones industrial average, Nasdaq, S&P 500, Nikkei 225, Hang Seng index, A-Share, SSE A-Share,FTSE 100,CAC 40, DAX and KOSPI index. Our finding reveals that conditional Value at Risk fitted ARMA(p,q)-GARCH(1,1) performs better than unconditional Value at Risk, and extreme value theory performed better than traditional method. Furthermore, our empirical result displays the performance of conditional Value at Risk which threshold level is decided by Hall1990 and Danielsson1997, indeed improve on model which threshold level is decided by past experience.

碩士
國立屏東商業技術學院
國際企業所
95
Foreign exchange has becoming more and more important, because of the trend of internationalization. Understanding the extreme behavior of foreign exchange rate will help manage foreign exchange rate risk. Therefore, this thesis investigates the extreme behavior of foreign exchange rate in G10 members by applying Extreme Value Theory (EVT) to the tail of the distribution of the daily rate of return of foreign exchange rate. In addition, we compare the EVT models by evaluating the forecasting performance of VaR. And we also investigate the influence of data dependence and choice of optimal threshold level on extreme value models. The empirical results show that compare to Normal distribution, daily return of foreign exchange rate is more Fat-tailed and asymmetric. This indicates that the normality assumption will lead underestimation of VaR. In backtesting, the conditional EVT models outperform the others, which imply that the dependence and conditional heteroscedasticity of time series should be accounted for when applying EVT. On the other hand, EVT models do not be affected when the threshold level changes. And we find parametric models generally outperform the non-parametrics model, especially the parametric model – GPD. GPD has the best and comprehensive performance both under uncondition and condition model. Moreover, the empirical results also show that GARCH model is adequate in forecasting VaR of lower confidence level (eg. 95%), however, at a higher confidence level (eg. 99.5%, 99.9%), EVT models provide more reliable VaR forecasting. And we can say that the application of EVT models in risk management is essential.

Department of Statistics
MSc (Statistics)
The ripple e ect of a stock market crash due to extremal dependence is a global issue with key attention and it is at the core of all modelling e orts in risk management. Two methods of extreme value theory (EVT) were used in this study to model equity risk and extremal dependence in the tails of stock market indices from four African emerging markets: South Africa, Nigeria, Kenya and Egypt. The rst is the \bivariate-threshold-excess model" and the second is the \point process approach". With regards to the univariate analysis, the rst nding in the study shows in descending hierarchy that volatility with persistence is highest in the South African market, followed by Egyptian market, then Nigerian market and lastly, the Kenyan equity market. In terms of risk hierarchy, the Egyptian EGX 30 market is the most risk-prone, followed by the South African JSE-ALSI market, then the Nigerian NIGALSH market and the least risky is the Kenyan NSE 20 market. It is therefore concluded that risk is not a brainchild of volatility in these markets. For the bivariate modelling, the extremal dependence ndings indicate that the African continent regional equity markets present a huge investment platform for investors and traders, and o er tremendous opportunity for portfolio diversi cation and investment synergies between markets. These synergistic opportunities are due to the markets being asymptotic (extremal) independent or (very) weak asymptotic dependent and negatively dependent. This outcome is consistent with the ndings of Alagidede (2008) who analysed these same markets using co-integration analysis. The bivariate-threshold-excess and point process models are appropriate for modelling the markets' risks. For modelling the extremal dependence however, given the same marginal threshold quantile, the point process has more access to the extreme observations due to its wider sphere of coverage than the bivariate-threshold-excess model.
NRF