Skewness and kurtosis are measures used to describe shape characteristics of distributions. In this thesis, we examine the interval estimates about the skewness and kurtosis by using jackknife empirical likelihood (JEL), adjusted JEL, extended JEL, traditional bootstrap, percentile bootstrap, and BCa bootstrap methods. The limiting distribution of the JEL ratio is the standard chi-squared distribution. The simulation study of this thesis makes a comparison of different methods in terms of the coverage probabilities and interval lengths under the standard normal distribution and exponential distribution. The proposed adjusted JEL and extended JEL perform better than the other methods. Finally we illustrate the proposed JEL methods and different bootstrap methods with three real data sets.

L’utilisation de câbles électriques et leurs longueurs dans certains systèmes électriques ont fortement augmenté au cours des dernières années. Or, la fiabilité de ces systèmes repose en partie sur la fiabilité des réseaux électriques. On constate en pratique qu’une part non négligeable des pannes et des dysfonctionnements de ces systèmes proviennent des défauts dans les liaisons filaires et non des équipements électriques. La connaissance de ces réseaux filaires et en particulier la détection de leurs défauts est donc importante. De nombreuses méthodes ont été développées pour tester l’état des câbles. Parmi ces méthodes on peut distinguer les méthodes de réflectométrie largement utilisées et facilement embarquables. Généralement ces méthodes sont très bien adaptées pour détecter et localiser les défauts francs mais les défauts non francs sont pratiquement transparents à ces méthodes car ils ont des conséquences électriques très faibles. Pour s’affranchir de ces limitations des améliorations en termes de mesure et traitement sont nécessaires. Dans cette thèse, trois nouvelles méthodes de diagnostic filaire ont été développées pour améliorer et faciliter la détection et la localisation de tous types de défauts filaires. Chacune des ces méthodes répond à un obstacle que nous avons rencontré pendant les trois années de recherche. Un premier obstacle concerne le phénomène de dispersion du signal dans les câbles qui rend la détection des défauts et du vieillissement des câbles très difficile. Un autre obstacle lié à la détection des défauts non-francs présente un enjeu actuel majeur du diagnostic filaire car leurs signatures sont très faibles et parfois noyées dans le bruit ou masquées par la proximité d’une autre impulsion d’amplitude plus importante. Les trois méthodes sont les suivantes : – La première méthode proposée, baptisée « corrélation adaptative » fournit un nouvel algorithme pour compenser la dispersion du signal. Elle permet de mieux localiser et mieux détecter les singularités sur des câbles de n’importe quelle longueur. – La deuxième méthode proposée, baptisée TRR (en anglais Time reversal Reflectometry) est basée sur le principe de la réflectométrie et du retournement temporel. Elle permet de caractériser le vieillissement des câbles électriques. – La troisième méthode proposée, baptisée RART (Réflectométrie associée à un processus de retournement temporel) est basée sur les principes de la réflectométrie et du retournement temporel et permet d’améliorer la détection des défauts électriques liés à une dégradation de l’isolant. Ces travaux de thèse ont montré les performances et la facilité de ces méthodes visant à assurer la sureté de fonctionnement des systèmes électriques que ce soit dans des moyens de transport, un bâtiment ou même des réseaux de communication. / The use of electric cables in electrical systems has been significantly increasing over the last decades. However, the reliability of these systems is partially based on the reliability of electrical networks. Current practices show that a significant number of failures and malfunctions of these systems come from faults in wired links and not from electrical devices. Therefore, the knowledge of the state of wire networks and particularly the detection of their faults is important. Several methods have been developed to test the status of cables. Among them, reflectometry methods are widely used and easily embeddable. Generally, these methods are appropriate to detect and locate hard faults but soft faults are virtually transparent to them because this kind of fault has very low electrical consequences. Improvements in measurement and treatment are necessary to overcome the limitations of these methods. In this respect, three new methods for wire diagnosis have been studied and developed to improve and ease the detection and location of soft wire faults. Each of these methods circumvents one or more of the barriers encountered during this research’s duration. First barrier, the phenomenon of signal dispersion in cables makes the detection of faults and of cable aging difficult or imprecise. Another barrier, the detection of soft faults, represents currently a major issue of wire diagnosis because the amplitude of soft faults signatures is very small and sometimes noisy or masked by the proximity of higher pulses. The three methods can briefly described as follows : – The first method, called "adaptive correlation", provides a new algorithm to compensate signal’s dispersion. It improves fulat’s location and the detection of singularities on cables regardless their lengths. – The second method, called TRR (Time Reversal Reflectometry), is based on the principle of reflectometry and time reversal. It allows the characterization of aging of electrical cables. – The third method, called RART (Reflectrometry combined with a time reversal process), is also based on the principle of reflectometry and time reversal. It improves the detection of electrical faults related to degradation of insulation. This research illustrates the efficiency and applicability of the proposed methods. It also demonstrates the potential of the proposed methods to improve safety in operation of electrical systems whether in transport, construction, or even communication networks.

[pt] Esta tese é composta por três estudos que têm em comum um papel destacado para choques de demanda na avaliação do risco de ações. O primeiro estudo avalia se ações que apresentam volume anormalmente alto têm um maior retorno nas semanas seguintes, sem necessariamente maior risco, como relatado anteriormente na literatura. Diferentemente de resultados precedentes, o estudo mostra que o risco sistemático pode explicar parcialmente o maior retorno de uma carteira formada com ações que apresentam volume anormalmente alto. Porém, também mostra que a correlação com o retorno do mercado não é suficiente para explicar o maior retorno da carteira de maior volume. O segundo estudo avalia se retornos acumulados afetam a resposta da volatilidade a choques de retorno. Foi verificado que sim. Este resultado é atribuído a viés comportamental que, atenuaria o impacto dos choques positivos, sobre a volatilidade, quando o retorno acumulado e corrente são negativos, mas amplificaria o impacto dos choques negativos, quando o retorno acumulado e corrente são negativos. O estudo apresenta evidência empírica que suporta esta explicação. O terceiro estudo, motivado por literatura recente que mostra a relevância da assimetria das distribuições de retornos de ações na avaliação de risco, procura identificar determinantes da assimetria. Além das variáveis explicativas identificadas pela literatura precedente, o estudo mostra que o ganho de capital e uma variável proposta como proxy para a freqüência de incorporação de novidades aos preços afetam a assimetria da distribuição de retornos de ações de maneira estatisticamente significativa e com os sinais conjecturados. / [en] This thesis consists of three essays which have in common the role of demand shocks in the evaluation of stocks risk. The first essay evaluates whether stocks that present abnormal high volume have a greater return in the following weeks, which is not necessarily linked to higher risk, as previously reported in financial literature. Contrary to previous results, the essay shows that systematic risk may partially explain the greater return of the portfolio formed with stocks that present abnormally high volume. However, it also shows that the correlation with market return is not sufficient to explain the greater return of the high volume portfolio. The second essay evaluates whether cumulative returns affect the response of volatility to return shocks. The result was affirmative, and attributed to behavioral bias, that attenuates the impact f positive shocks, when cumulative and current return are positive, and amplifies the impact of negative shocks, when cumulative and current return are negative. The essay also provides empirical evidence supporting this explanation. The third essay, motivated by recent literature that shows the relevance of skewness of returns distribution to risk evaluation, aims the identification of skewness determinants. Besides explanatory variables identified by previous literature, the essay shows that capital gain and a variable proposed as proxy for the frequency of information incorporation into prices affect the skewness of stocks returns distribution, with statistical significance and conjectured signs.

Liu (1996) discussed a class of robust normal/independent distributions which contains a group of thick-tailed cases. In this article, we develop a skewed version of these distributions in the multivariate setting, and we call them multivariate skew normal/independent distributions. We derive several useful properties for them. The main virtue of the members of this family is that they are easy to simulate and lend themselves to an EM-type algorithm for maximum likelihood estimation. For two multivariate models of practical interest, the EM-type algorithm has been discussed with emphasis on the skew-t, the skew-slash, and the contaminated skew-normal distributions. Results obtained from simulated and two real data sets are also reported. / Liu (1996) discute una clase de distribuciones robustas a las que apela como normal/independiente, y que contiene un grupo de distribuciones de colas pesadas. En este artículo desarrollamos una versión asimétrica de tales distribuciones en un escenario multivariado, a las que llamaremos distruciones normales asimétricas independientes multivariadas. Para tales distribuciones derivamos algunas propiedades. La principal virtud de los miembros de esta familia es que son fáciles de simular y se prestan a un algoritmo de tipo EM para realizar estimaciones de máxima verosimilitud de sus parámetros. Para dos modelos multivariados de interés práctico se discute el algoritmo EM con énfasis en las distribuciones t-asimétrica, slash asimétrica y normal asimétrica contaminada. Los resultados obtenidos a partir de simulaciones y de dos conjuntos de datos reales son reportados.

Implied volatility obtained from market option prices is widely regarded as an efficient predictor of future realised volatility. Implied volatility can be thought of as market's expectation of future realised volatility. We distinguish between volatility-changing events with respect to expectations - scheduled events (such as information releases) and unscheduled events. We propose a method of testing the information content of option-implied risk-neutral moments prior to volatility-changing events. Using the method introduced by Bakshi, Kapadia & Madan (2003) we extract implied volatility, skewness and kurtosis from S&P 500 options market prices and apply the proposed method in four case studies. Two are concerned with scheduled events - United Kingdom European Union membership referendum, 2016 and United States presidential election, 2016, two are concerned with unscheduled events - flash crash of August 24, 2015 and flash crash of October 15, 2014. Implied volatility indicates a rise in future realised volatility prior to both scheduled events. We find a significant rise in implied kurtosis during the last three days prior to the presidential election of 2016. Prior to unscheduled events, we find no evidence of implied moments indicating a rise in future realised volatility.

[pt] A avaliação dos projetos de investimentos é uma tarefa difícil para muitas empresas, especialmente para aqueles cujo fluxo de caixa depende dos preços das commodities, já que o nível de incerteza nos preços tem um alto impacto na determinação do momento adequado para o investimento. Os métodos de avaliação tradicionais, que não levam em consideração a flexibilidade gerencial nem a modelagem da incerteza do projeto, podem levar a decisões não ótimas. Esta pesquisa desenvolve um modelo que considera estas variáveis, usando árvores binomiais implícitas ajustados por outros indicadores de risco, como assimetria e curtose da rentabilidade do projeto. O nível de incerteza pode não só ser medido pela volatilidade do retorno do projeto, mas também pela probabilidade de se obter um resultado baixo ou negativo no projeto. A magnitude dessa probabilidade poderia ser a avaliada conhecendo-se o valor da assimetria e curtose do retorno do projeto. Para modelar o comportamento de um projeto, esta dissertação apresenta dois tipos de árvores binomiais implícitas, recombinantes e não recombinante. Cada árvore tem sua própria abordagem específica para determinar o valor do projeto, incluindo opções. Um caso aplicado é apresentado considerando uma empresa de mineração. Os resultados sugerem que o nível de assimetria contribui para uma melhor avaliação do risco do projeto, que combinado com a metodologia de opções reais captura melhor o valor das flexibilidades do projeto; o que é uma importante contribuição do modelo proposto nesta dissertação. / [en] Valuation of capital investment projects is a difficult task for many companies, especially for those whose cash flows depend on commodity prices. The level of uncertainty in commodity prices has a significant impact in determining the proper timing for an investment. Traditional valuation methods, which do not take into account managerial flexibility or project uncertainty modeling can lead to non-optimal decisions. This research develops a dynamic model that considers these variables, and uses implied binomial trees adjusted by other indicators of risk, such as project return s skewness and kurtosis. The level of uncertainty can not only be measured by the project return s volatility, but also by how probable is the occurrence of a low or negative result in the project. The magnitude of this probability could be assessed by knowing the project return s skewness and kurtosis. To model the project s behavior, this dissertation presents two kinds of implied binomial trees, recombining and non-recombining trees. Each tree has its own specific approach to determining the value of the project, including options or managerial flexibility. An applied case is presented considering a mining project. The results suggest that the level of skewness helps to have a better measure of project risk, which combined with the real option approach, allows capturing the value of project managerial flexibilities; which is an important contribution of the proposed model in this dissertation.

La simulation des grandes échelles (SGE ou LES pour large eddy simulation) commence à être très utilisée dans l’industrie. Par résolution directe des structures turbulents de grande tailles, le calcul LES est capable de calculer le bruit générée par la voilure ou de prédire avec précision le décollement de coin dans une configuration très simplifiée du compresseur. L’un des problèmes les plus importants pour effectuer un calcul LES est de fournir des conditions d’entrée avec des champs turbulents.Pour une approche hybride RANS/LES (RANS pour Reynolds Averaged Navier-Stokes), les conditions d’entrée turbulentes pour un calcul LES sont générées à l’aide des solutions fournies par le calcul RANS en amont. Il existe plusieurs méthodes pour générer les conditions d’entrée pour LES. Elles peuvent principalement être classées en deux catégories : 1) simulation avec pré-calcul ; 2) la méthode de turbulence synthétique. La simulation avec pré-calcul consiste à effectuer un calcul LES indépendant pour générer un champ turbulent comme conditions d’entrée pour alimenter le calcul principal. Cette méthode peut obtenir des turbulences de haute qualité, mais elle augmente considérablement le temps de calcul et le stockage des données. Le champ turbulent généré par la méthode de turbulence synthétique exige une « distance de adaptation », pendante laquelle le champ turbulent devient pleinement développé. L’objectif principal pour améliorer ce genre de méthodes est donc de diminuer cette distance nécessaire.Dans cette thèse, la méthode de vortex, qui est une approche de turbulence synthétique, est présentée et améliorée. A travers des expériences numériques, les paramètres de la méthode de vortex améliorée sont systématiquement optimisés. L’application à l’écoulement en canal plan et à couche limite en plaque plane, montrent que la méthode de vortex améliorée génère de manière efficace pour fournir des conditions d’entrée pour LES. Dans le cas de l’écoulement en canal plan, la distance d’adaptation nécessaire pour le rétablissement de la turbulence est d’environ 6 fois la demi-hauteur du canal. Pour le cas de l’écoulement en plaque plane, cette distance est environ 21 fois l’épaisseur de la couche limite. Enfin, dans le but de qualifier la turbulence obtenue par des calculs LES, nous utilisons les coefficients de dissymétrie des dérivées des fluctuations de vitesse, et, nous les introduisons comme un nouveau critère pour la qualité de LES. / Large eddy simulation is becoming an important numerical tool in industry recently. Resolving large scale turbulent motions directly, LES is capable to compute the aeroacoustic noise generated by the airfoil or to precisely capture the corner separation in a linear compressor cascade. The main challenge to perform a LES calculation is to prescribe a realistic unsteady inflow field. For hybrid RANS/LES approaches, inflow conditions for downstream LES region must be generated from the upstream RANS solutions. There exist several methods to generate inflow conditions for LES. They can mainly be divided into two categories: 1) Precursor simulation; 2) Synthetic turbulence methods. Precursor simulation requires to run a separate calculation to generate a turbulent ow or a database to feed the main computation. This kind of methods can generate high quality turbulence. However, it requires heavy extra computing load. Synthetic turbulence methods consist in generating a fluctuating velocity field, and within a short “adaptation distance”, the field get fully developed. So main goal of synthetic turbulence methods is to decrease the required adaptation distance. The vortex method which is a synthetic turbulence method is presented and improved here. Parameters of the improved vortex method are optimized systematically with a series of calculations in this thesis. Applications on channel and flat-plate flows show that the improved vortex method is effective in generating the LES inflow conditions. The adaptation distance required for turbulence recovery is about 6 times the half channel height for channel flow, and 21 times the boundary-layer thickness (at the inlet of vortex) for at-plate ow. The velocity-derivative skewness is used to qualify the generated turbulence, and is introduced as a new criterion of LES calculation.

No description available.

In this thesis, we present the use of logistic regression method to develop a credit scoring modelusing the raw data of 4447 customers of a bank. The data of customers is collected under 14independent explanatory variables and 1 default indicator. The objective of this thesis is toidentify optimal coefficients. In order to clean data, the raw data set was put through variousdata calibration techniques such as Kurtosis, Skewness, Winsorization to eliminate outliers.On this winsorized dataset, LOGIT analysis is applied in two rounds with multiple statisticaltests. These tests aim to estimate the significance of each independent variable and modelfitness. The optimal coefficients can be used to obtain the credit scores for new customers witha new data set and rank them according to their credit risk.

This paper examines the extent to which thin markets in conjunction with tax induced kinks in the budget constraint cause consumer demand to be skewed. To illustrate the principles I focus on the demand for owner-occupied housing. Housing units are indivisible and heterogeneous while tastes for housing are at least partly idiosyncratic, causing housing markets to be thin. In addition, prior to 1998, capital gains tax provisions introduced a sharp kink in the budget constraint of existing owner-occupiers in search of a new home: previous homeowners under age 55 paid no capital gains tax if they bought up, but were subject to capital gains tax if they bought down. I first characterize the economic conditions under which households err on the up or down side when choosing a home in the presence of a thin market and a kinked budget constraint. I then specify an empirical model that takes such effects into account. Results based on Monte Carlo experiments indicate that failing to allow for skewness in the demand for housing leads to biased estimates of the elasticities of demand when such skewness is actually present. In addition, estimates based on American Housing Survey data suggest that such bias is substantial: controlling for skewness reduces the price elasticity of demand among previous owner-occupiers from 1.6 to 0.3. Moreover, 58% of previous homeowners err on the up while only 42% err on the down side. Thus, housing demand is skewed. / Ph. D.