The main topic of the thesis is a discussion of the contraction of esse (-st < est , -'s < -es), a widespread phenomenon in Latin comedy, generally known as ‘prodelision’. The thesis collects evidence for contracted forms from a large variety of sources (e.g. inscriptions, Sabellian languages and ancient grammarians) and analyses their transmission in the manuscript tradition of Latin authors. Contracted forms appear to be widespread but are neglected by or puzzling to editors. They are not misspellings, abbreviations, or representations of a sandhi phenomenon related to elision (‘prodelision’), but are clitic forms of the verb esse, showing phonetic reduction. The thesis discusses their linguistic significance, relying in particular on analysis of their behaviour in Terence: auxiliary or copula esse is clitic and its standard position seems to be clause-final, attached to the participle or predicate noun or adjective. Second, the thesis analyses another complex phenomenon of early Latin, the prosodical omission of final -s before a consonant (sigmatic ecthlipsis), relying on a database of all Terence's lines potentially involving it. Metrical evidence appears to be limited to cases in which a participle or predicate noun or adjective precedes a form of esse beginning with s-, thus confirming the prosodical distinctiveness of such a sequence. The evidence of sigmatic ecthlipsis thus ties in closely with that of contraction. The findings of the thesis have philological and editorial implications. First, they throw light on orthographic, phonetic, prosodic, and syntactic aspects of the verb ‘to be’ in Latin. It has a strong bond with the participle and predicate noun or adjective and may either be reduced phonetically in combination with such hosts, or participate in a simplification of -ss- which it may share with its host. Second, transmitted contracted spellings or readings betraying a misunderstanding of them should be given full consideration by editors and not disregarded. Contracted spellings might be restored in other conditions since manuscripts are often not trustworthy when transmitting uncontracted forms. Contracted forms may be stylistically motivated. They occur in comedy particularly in spoken metres, and in late Latin become a spelling archaism. Finally, sigmatic ecthlipsis in Terence occurs for certain only in restricted conditions: the editorial practice of printing it wherever it is possible even if not necessary is misleading.

Les futures normes de solvabilité pour l’industrie de l’assurance, Solvabilité 2, ont pour buts d’améliorer la gestion des risques au travers de l’identification de différentes classes et modules de risque, et en autorisant les compagnies à utiliser des modèles internes pour estimer leur capital réglementaire. La formule standard définit ce capital comme étant égal à une VaR à 99.5% sur un horizon d’un an pour chaque module de risque. Puis, à chaque niveau de consolidation intermédiaire, les différentes VaR sont agrégées au travers d’une matrice de corrélation. Plusieurs problèmes apparaissent avec cette méthode : – Le régulateur utilise le terme de “VaR” sans communiquer de distributions marginales ni globale. Cette mesure de risque multi-variée n’est pertinente que si chaque risque suit une distribution normale. – L’horizon temporel à un an ne correspond pas à celui des engagements d’une compagnie d’assurance, et pose des problèmes d`es lors qu’il faut déterminer la fréquence de mises à jour des modèles internes. – La structure de dépendance proposée par la formule standard ne correspond pas à celle habituellement mise en place par les compagnies et est difficilement utilisable dans un modèle interne. La première partie présentera en détail les points clés de la réforme et donnera des axes de réflexion sur son application dans la gestion des risques. Dans une deuxième partie, il sera montré que cette mesure de risque multi-variée ne satisfait pas aux principaux axiomes d’une mesure de risque. De plus, elle ne permet pas de comparer les exigences de capital entre compagnies, puisqu’elle n’est pas universelle. La troisième partie démontrera que pour évaluer un capital à un point intermédiaire avant l’échéance, une mesure de risque doit pouvoir s’ajuster à différentes périodes, et donc être multi-périodique. Enfin, la quatrième partie mettra l’accent sur une alternative à la matrice de corrélation pour modéliser la dépendance, à savoir les copules. / The new rules of solvency for the insurance industry, Solvency II, aim to improve the risk management in the insurance industry by identifying different classes / modules of risk, and by allowing insurance companies to use an internal model to estimate their capital. The standard formula sets the capital requirement at a VaR of 99.5% level for a one year horizon for each sub risk module. Then at each consolidation level, the different VaR are aggregated through a correlation matrix. Some problems may appear with this method : – The regulator uses “VaR” term while he provides neither marginal distributions nor the global one. This multivariate risk measure is relevant only if each risk follows a normal distribution. – This short term horizon does not match the time horizon of the liabilities of an insurance company and leads to some problems in updating the capital requirement during the year. – The dependance structure given in the standard formula does not correspond to a practical one, and cannot be used in an internal model. The first part will present a detailed discussion about the reform and give some example of its application from risk management’s point of view. In the second part, it will be establish that this multivariate risk measure does not satisfy the main axioms that a risk measure should fulfill. With this approach, there is not uniqueness among the insurance companies, so the solvency capital requirement cannot be compared across the industry. The third part will demonstrate that a risk measure which adjusts to different periods should be used to evaluate the capital at a point in time, a multiperiod risk measure. At last, the fourth part will emphasize on an alternative to the correlation matrix to aggregate risks, the copula.

<p>Bayesian methods offer a flexible and convenient probabilistic learning framework to extract interpretable knowledge from complex and structured data. Such methods can characterize dependencies among multiple levels of hidden variables and share statistical strength across heterogeneous sources. In the first part of this dissertation, we develop two dependent variational inference methods for full posterior approximation in non-conjugate Bayesian models through hierarchical mixture- and copula-based variational proposals, respectively. The proposed methods move beyond the widely used factorized approximation to the posterior and provide generic applicability to a broad class of probabilistic models with minimal model-specific derivations. In the second part of this dissertation, we design probabilistic graphical models to accommodate multimodal data, describe dynamical behaviors and account for task heterogeneity. In particular, the sparse latent factor model is able to reveal common low-dimensional structures from high-dimensional data. We demonstrate the effectiveness of the proposed statistical learning methods on both synthetic and real-world data.</p> / Dissertation

The goal of this project is to price multi-name credit derivatives using a copula approach. The properties and advantage copula functions have to other traditional methods are carefully evaluated. Monte Carlo simulations are studied and performed to obtain numerical results for copula functions with explicit and implicit forms. A model was developed to price a basic form of a first-to-default basket using different copula functions. The outcomes are analyzed and comparisons are carried out.

Em vista do sistema de Caso/concordância de Chomsky (2000; 2001) baseado em Agree, bem como na proposta minimalista para a ergatividade de Aldridge (2004; 2008; 2012), apresentamos no capítulo 1 uma análise para as relações de Caso e concordância em Karitiana à luz do seu padrão de concordância ergativo-absolutivo e das ordens de constituinte que a língua exibe. Propomos que T valora Caso absolutivo no objeto e no sujeito intransitivo em orações matriz, ao passo em que o núcleo n valora genitivo nestes argumentos em orações subordinadas. O núcleo v, por sua vez, atribui Caso ergativo inerente ao sujeito transitivo. Propomos, também, que o v transitivo possua um traço EPP responsável por alçar o objeto para uma posição de especificador extra de vP acima do sujeito, o que explicaria a ordenação atestada em sentenças assertivas VOS e subordinadas OSV. A disposição SVO em declarativas, em que o sujeito sobe mesmo estando abaixo do objeto em Spec,vP, é explicada em termos de uma restrição imposta sobre o objeto pela Condição de Atividade (CHOMSKY, 2001). No capítulo 2 discutimos de que maneira as relações predicativas são sintaticamente capturadas nas construções de cópula do Karitiana. Adotamos a teoria da Sintaxe da Predicação de den Dikken (2006), em que sujeito e predicado estabelecem uma relação de predicação no domínio da projeção de um núcleo relator em uma configuração assimétrica e não direcional. Também lançamos mão da proposta em Hale & Keyser (2002) de que a projeção de um especificador depende de um núcleo verbal selecionar um elemento predicador como um complemento. Apresentamos, assim, uma análise para as construções de cópula de predicação verbal e adjetival (i.e. [NP cóp i-Pred-t/ø]). Em nossa proposta o pronome i- prefixado ao adjetivo e verbo corresponde a um pro-predicate (MORO, 1997), um place holder do predicado lexical. Tal pro-predicate é o verdadeiro predicador da construção de cópula em que ele aparece, de modo que o adjetivo e o verbo entram como um adjunto na estrutura. Apresentamos, também, uma derivação para as construções de inversão de predicado, nas quais o i- está ausente. O morfema -t/ø sufixado ao predicado corresponde a um bloqueador de projeção, sendo uma maneira que a língua possui de contornar o problema de rotulação que emerge em certas estruturas de adjunção. Por fim, explicamos o comportamento distinto das construções de predicação nominal (i.e. [ NP cóp NP-t/ø]) argumentando que elas não se submetem às restrições da sintaxe da predicação. Finalmente, no capítulo 3 argumentamos que o apagamento da cópula em Karitiana nas sentenças declarativas de tempo não futuro é um caso de elipse do tipo sluicing (MERCHANT, 2001) que não precisa satisfazer condições de identidade para ser licenciada, propondo que o efeito copula drop decorre de uma restrição baseada em fases (CHOMSKY, 2001; BOKOVI, 2014). / In light of the chomskyan Case/agreement system (2000; 2001), and also Aldridges (2004; 2008; 2012) minimalist approach to ergativity, in chapter 1 we present an account for Case and agreement relations in Karitiana in view of its ergative-absolutive agreement pattern and the constituent order the language exhibits. We propose T values absolutive Case on both the object and the intransitive subject within matrix clauses, and n head values genitive on these arguments within subordinate clauses. We also propose v assigns ergative as an inherent Case to the transitive subject. In addition, transitive v bears an EPP feature responsible for dragging the object to an outer Spec,vP position above the subject, which accounts for the VOS assertives and OSV subordinates word order. SVO declaratives, in which the subject is raised even though the object is arguably higher up in the tree, is derived due to a restriction the Activity Condition (CHOMSKY, 2001) imposes on the object. In chapter 2 we discuss how predication relations are syntactically captured in Karitiana copular constructions. We adopt den Dikkens (2006) Predication theory, in which the subject and the predicate establish a relation in the minimal domain of the projection headed by a relator in a asymmetrical and non directional configuration. We also adopt Hale & Keysers (2002) proposal in which the specifier position is projected only when a verbal head selects for a predicate element as its complement. Then, we present an analysis for verbal and adjectival predication in Karitiana copular constructions (i.e. [NP cop i-A/V-t/ø]). In our approach the pronoun i- prefixed to the adjective and the verb is a pro-predicate (MORO, 1997), a place holder for the lexical predicate. Such element is the real predicate in the copula constructions it appears, and the adjective and the verb are adjoined to the structure. We present, as well, a derivation for predicate inversion constructions, which lack i-. The morpheme -t/ø suffixed to the predicate corresponds to a projection blocker, and it is a way the language alleviates a projection problem caused by certain adjunction structures. To conclude, we explain the distinctive behaviour of nominal predicate constructions (i.e. [ NP cop NP-t/ø]) arguing they are not submitted to restrictions of the Syntax of Predication. Finally, in chapter 3 we propose a phase-based (CHOMSKY, 2001; BOKOVI, 2014) account of the copula drop phenomenon in Karitiana declarative sentences (non future tense), arguing it to be an instance of sluicing (MERCHANT, 2001), a type of ellipsis, and that it does not need to satisfy identity conditions to be licensed.

Due to turbulence in the financial market throughout history, stress testing has become a growing part of the risk analysis performed by clearing houses. Events connected to previous crises have increased the demand for prudent risk exposure, and in this thesis we investigate regulators view on how CCPs should construct risk scenarios to meet best practice for stress testing their members’ composite portfolios. A method based on multivariate t-distributions and copula-transformations applied to historical time series data, is proposed for constructing an independent scenario generator which should be used as a compliment to other, more knowledge-based methods. The method was implemented in Matlab to test the theory in practice, and experiments were setup for pure stock portfolios as well as for derivative based portfolios. Backtests were then carried out to validate the underlying theory on historical data spanning 25 years in total. Results show that the method proposed in this thesis indeed has the potential to be a useful approach for creating stress scenarios. Its ability to render specific levels of plausibility seems to show a sufficient level of consistency with real life data, and further research is thereby justified.

Esta tese tem por objetivo descrever e analisar, com base na Teoria Gerativa, as construções semiclivadas e pseudoclivadas no português brasileiro. Reconhecemos, para ambas as construções, a necessidade de uma análise dual, que identifica dois tipos. Quanto às semiclivadas, um tipo (e este é o grande grupo) recebe uma análise monoracional e independente das pseudoclivadas, são as chamadas semiclivadas verdadeiras, e o outro, bem delimitado e restrito, recebe uma análise bioracional, como uma versão reduzida das pseudoclivadas. Em nossa proposta de uma análise monoracional para a sintaxe das semiclivadas verdadeiras, a cópula, que imediatamente precede o foco, é a realização de uma categoria funcional envolvida no estabelecimento de relações de predicação, um RELATOR ou LINKER, seguindo a sintaxe da predicação desenvolvida em Den Dikken (2006a). Explorando a aplicação de Inversão de Predicado, com suas implicações, como a emergência de uma cópula, o congelamento do sujeito da predicação (que fica in situ) para manipulações sintáticas, e que recebe, invariavelmente, a interpretação de foco nas interfaces, alargamos o tratamento conferido à sintaxe da predicação, tratando inclusive a relação \'objeto de\' como uma relação essencialmente predicacional. Por sua vez, quanto às pseudoclivadas, os dois tipos reconhecidos recebem ambos análises bioracionais (contra análises monoracionais e de reconstrução). A proposta de uma análise dual para as pseudoclivadas, em tipo A e tipo B, na esteira de Den Dikken, Meinunger & Wilder (2000), se fundamenta na heterogeneidade dos dados em nossa língua - sobretudo com relação às diferenças entre os dois padrões dessas construções (com oração-wh inicial e final) e ao tipo de oração-wh que pode ocorrer em cada qual (interrogativa ou relativa livre) - que resistem a um tratamento único, inteiramente homogêneo. Essa análise dual tem alcance interlinguístico, parcial ou total, de acordo com as características das línguas. Procuramos mostrar como orações-wh diferentes estão relacionadas a padrões diferentes de pseudoclivadas, cada qual com uma estrutura sintática particular, bem como algumas decorrências que delas surgem e que contrastam entre si, conforme previsto. Mostramos, ainda, dentre os polêmicos efeitos de conectividade que as pseudoclivadas podem exibir, quais merecem tratamento sintático e quais não. Conectividades envolvendo anáfora e pronomes ligados são melhor tratadas sob uma perspectiva semântica, haja vista sua ocorrência mesmo em copulares especificacionais não-clivadas, independentemente da ordem entre os termos pré e pós-cópula, e para as quais mesmo um tratamento via reconstrução (última chance de um tratamento ainda sintático) é, às vezes, impossível. Crucialmente, as conectividades envolvendo IPN (itens de polaridade negativa) e Caso são as distintivas para a análise das pseudoclivadas em dois tipos. As pseudoclivadas do tipo A recebem uma análise bioracional com elipse, na qual a oração-wh é uma interrogativa e o contrapeso, uma resposta sentencial completa para essa pergunta, sujeita à elipse preferencial do material repetido. Assim, em analogia aos question-answer pairs, essas pseudoclivadas são chamadas de self-answering questions, com uma estrutura \'tópico-comentário\', cuja ordem é rígida, resultando somente no padrão com oração-wh inicial (pergunta&lt;resposta). Já as pseudoclivadas do tipo B, por sua vez, têm uma estrutura predicacional como base, uma small clause. Elas recebem uma análise também bioracional, mas do tipo WYSIWYG (what you see is what you get), em que a oração-wh, aqui uma relativa livre, é o predicado, e o XP, seu sujeito, necessariamente o foco da sentença. A relativa livre, na qualidade de um predicado nominal, está sujeita à Inversão de Predicado; consequentemente, a ordem das pseudoclivadas do tipo B é mais flexível, produzindo tanto o padrão com oração-wh inicial, quanto final. Por fim, resgatamos o segundo tipo de semiclivadas, o pequeno grupo restrito que é uma versão reduzida das pseudoclivadas, como evidenciado pelos fatos relativos à concordância do verbo lexical. A ocorrência dessas semiclivadas é limitada às semiclivadas de sujeito. Em face de dois tipos de pseudoclivadas, mostramos que somente as do tipo A permitem \'redução\'. Colaboram para isso a rigidez na ordem, característica das semiclivadas e apenas das pseudoclivadas do tipo A, e a possibilidade de omissão do pronome wh (\'wh-drop\'), possível em interrogativas (como uma opção disponível na Gramática Universal), mas nunca em relativas livres. Sendo essas pseudoclivadas reduzidas restritas às de sujeito, seu wh nulo (ocorrência de \'wh-drop\') é reanalisado como prowh nas línguas românicas que dispõem dessas construções. / This dissertation aims to describe and analyse semiclefts and pseudoclefts constructions in Brazilian Portuguese, with the background of Generative Theory. Given the heterogeneity of the facts analysed, we propose an approach that divides both constructions in two types. Regarding the semiclefts, one type (and this is the big group) receives a monoclausal analysis, independent of pseudoclefts, which are called true semiclefts, whereas the other, well delimited and restricted, receives a biclausal analysis, as a reduced version of pseudoclefts. In our proposal of a monoclausal analysis to the syntax of true semiclefts, the copula, which immediately precedes the focus, is the spell out of a functional category involved in the establishment of predication relations, either a RELATOR or a LINKER, following the syntax of predication developed in Den Dikken (2006a). Exploring the application of Predicate Inversion, with its implications, such as the emergence of a copula, the freezing of the subject of predication to syntactic manipulations which stays in situ and receives, invariably, the focus interpretation at the interfaces, we enlarge the approach given to the syntax of predication, treating also the \'object of\' relation as an essentially predicational relation. As for the pseudoclefts, the two types recognized receive both a biclausal analysis (contra monoclausal and reconstruction analyses). The proposal of a dual analysis to pseudoclefts, in type A and type B, a la Den Dikken, Meinunger & Wilder (2000), is based on the heterogeneity of the data in our language - especially regarding the differences attested between the two patterns of these constructions (with wh-clause initial or final) and regarding the type of wh-clause that can occur in each of them (an interrogative or a free relative) - data that resist a single fully homogeneous treatment. This dual analysis has crosslinguistic (partial or total) scope, according to the characteristics of the languages. We tried to show how different wh-clauses are related to different patterns of pseudoclefts, each of them with its particular syntactic structure, as well as some consequences that they give rise to and that contrast with each other, as predicted. We also claim that, among the polemic connectivity effects that pseudoclefts can have, some deserve a syntactic account, whereas others do not. Connectivity effects related to anaphors and bound pronouns are better accounted for by a semantic account, given their occurrence even in simple (non-cleft) specificational copular sentences, regardless of the order between the pre and postcopular terms, and for which even an analysis via reconstruction (last chance to still have a syntactic treatment) is sometimes impossible. Crucially, connectivity effects related to NPI (negative polarity items) and Case are the relevant ones to distinguish the pseudoclefts in two types. Type A pseudoclefts receive a biclausal analysis with ellipsis, in which the wh-clause is an interrogative and the counterweight is a full sentence, the complete answer to the question, subject to (strongly favoured) ellipsis of the repeated part. In analogy to the question-answer pairs, these pseudoclefts are called self-answering questions, with a \'topic-comment\' structure, whose order is rigid, resulting only in the wh-clause-initial pattern (question&lt;answer). On the other hand, type B pseudoclefts are based on a predicational structure, a small clause. They also receive a biclausal analysis, but one of the WYSIWYG (what you see is what you get) kind, in which the wh-clause, here a free relative, is the predicate, and the XP, its subject, necessarily the focus of the sentence. The free relative, as a predicate nominal, is subject to Predicate Inversion; consequently, the order of type B pseudoclefts is more flexible, resulting in both patterns, wh-clause-initial or final. Finally, we go back to semiclefts, more specifically, to its the second type, the limited group that is a reduced version of pseudoclefts, as indicated by the facts related to the agreement in the lexical verb. The occurrence of such semiclefts is restricted to the subject semiclefts. In face of two types of pseudoclefts, we show that only type A allow \'reduction\'. Factors that contribute to this are the rigid order, characteristic of semiclefts and only of type A pseudoclefts, and the possibility of omission of the wh pronoun (\'wh-drop\'), possible in interrogatives (as an available option in Universal Grammar), but never in free relatives. Since these reduced pseudoclefts are limited to the subject ones, their null wh (an instance of \'wh-drop\') is reanalysed as a prowh in the romance languages that have these constructions.

Modelos de mistura e de não mistura em longa duracão, são aplicados na analise de dados de sobrevivência quando uma parcela de indivduos não são suscetíveis ao evento de interesse. Diferentes modelos estatsticos são propostos para analisar dados de sobrevivência na presenca de fracão de cura. Nesta tese, e proposto o uso de novos modelos. Sob o ponto de vista univariado, inicialmente e considerado o caso em que os dados de sobrevivênciaa seguem distribuicão Burr XII com três parâmetros, no qual inclui o modelo de mistura para a distribuicão Weibull como caso particular. Um modelo de sobrevivência geral e estudado considerando a situacão em que os parâmtreos de locacão e forma dessa distribuicão dependem de covariaveis. Ainda considerando o caso univariado, um estudo da distribuicãoo exponencial exponenciada com dois parâmetros e realizado. Essa distribuicão, tambem conhecida como distribuicão exponencial generalizada, e um caso particular da distribuicão Weibull exponenciada, introduzida por Mudholkar e Srivastava (1993). Um modelo de sobrevivência geral tambem e estudado, nesse caso considera-se a situacão em que os parâmetros de escala, forma e de fracão de cura da distribuicão exponencial exponenciada dependem de covariaveis. Um terceiro estudo univariado considera a distribuicão Weibull na presenca de fracão de cura, dados censurados e covariaveis. Nesse caso, dois modelos são estudados: modelo de mistura e modelo de não mistura. Quando dois tempos de sobrevivência distintos estão associados a cada unidade amostral (caso bivariado), na analise dos dados e possvel utilizar algumas distribuicões bivariadas: em especial a distribuicão exponencial bivariada de Block e Basu. As estimativas dos parâmetros da distribuicão exponencial bivariada de Block e Basu na presenca de fracão de cura e covariaveis são obtidas. Sob o ponto de vista bivariado tambem sera considerado o caso da distribuicão Weibull bivariada derivada de função copula na presenca de fração ao de cura, dados censurados e covariaveis. Duas funcões copulas são exploradas: a funcão copula Farlie-Gumbel-Morgenstern (FGM) e a funcão copula Gumbel. Procedimentos classicos e Bayesianos são utilizados para obter estimadores pontuais e intervalares dos parâmetros desconhecidos. Para vericar a utilidade e o comportamento dos modelos, alguns conjuntos de dados na area medica são analisados. / Mixture and non-mixture lifetime models are applied to analyze survival data when some individuals may never experience the event of interest. Dierent statistical models are proposed to analyze survival data in the presence of cure fraction. In this thesis, we propose the use of new models. From the univariate case, we consider that the lifetime data have a three-parameter Burr XII distribution, which includes the popular Weibull mixture model as a special case. We consider a general survival model where the scale and shape parameters of the Burr XII distribution depends on covariates. Also considering the univariate case the two-parameters exponentiated exponential distribution is used. The two-parameter exponentiated exponential or the generalized exponential distribution is a particular member of the exponentiated Weibull distribution introduced by Mudholkar and Srivastava (1993). We also consider in this case a general survival model where the scale, shape and cured fraction parameters of the exponentiated exponential distribution depends on covariates. We also introduce the univariate Weibull distributions in presence of cure fraction, censored data and covariates. Two models are explored in this case: the mixture model and non-mixture model. When we have two lifetimes associated with each unit (bivariate data), we can use some bivariate distributions: as special case the Block and Basu bivariate lifetime distribution. We also presents estimates for the parameters included in Block and Basu bivariate lifetime distribution in presence of covariates and cure fraction, applied to analyze survival data when some individuals may never experience the event of interest and two lifetimes are associated with each unit. We also consider in bivariate case the bivariate Weibull distributions derived from copula functions in presence of cure fraction, censored data and covariates. Two copula functions are explored in this paper: the Farlie-Gumbel-Morgenstern copula (FGM) and the Gumbel copula. Classical and Bayesian procedures are used to get point and condence intervals of the unknown parameters. Illustrations of the proposed methodologies are given considering medicals data sets.

Modelos de mistura e de não mistura em longa duracão, são aplicados na analise de dados de sobrevivência quando uma parcela de indivduos não são suscetíveis ao evento de interesse. Diferentes modelos estatsticos são propostos para analisar dados de sobrevivência na presenca de fracão de cura. Nesta tese, e proposto o uso de novos modelos. Sob o ponto de vista univariado, inicialmente e considerado o caso em que os dados de sobrevivênciaa seguem distribuicão Burr XII com três parâmetros, no qual inclui o modelo de mistura para a distribuicão Weibull como caso particular. Um modelo de sobrevivência geral e estudado considerando a situacão em que os parâmtreos de locacão e forma dessa distribuicão dependem de covariaveis. Ainda considerando o caso univariado, um estudo da distribuicãoo exponencial exponenciada com dois parâmetros e realizado. Essa distribuicão, tambem conhecida como distribuicão exponencial generalizada, e um caso particular da distribuicão Weibull exponenciada, introduzida por Mudholkar e Srivastava (1993). Um modelo de sobrevivência geral tambem e estudado, nesse caso considera-se a situacão em que os parâmetros de escala, forma e de fracão de cura da distribuicão exponencial exponenciada dependem de covariaveis. Um terceiro estudo univariado considera a distribuicão Weibull na presenca de fracão de cura, dados censurados e covariaveis. Nesse caso, dois modelos são estudados: modelo de mistura e modelo de não mistura. Quando dois tempos de sobrevivência distintos estão associados a cada unidade amostral (caso bivariado), na analise dos dados e possvel utilizar algumas distribuicões bivariadas: em especial a distribuicão exponencial bivariada de Block e Basu. As estimativas dos parâmetros da distribuicão exponencial bivariada de Block e Basu na presenca de fracão de cura e covariaveis são obtidas. Sob o ponto de vista bivariado tambem sera considerado o caso da distribuicão Weibull bivariada derivada de função copula na presenca de fração ao de cura, dados censurados e covariaveis. Duas funcões copulas são exploradas: a funcão copula Farlie-Gumbel-Morgenstern (FGM) e a funcão copula Gumbel. Procedimentos classicos e Bayesianos são utilizados para obter estimadores pontuais e intervalares dos parâmetros desconhecidos. Para vericar a utilidade e o comportamento dos modelos, alguns conjuntos de dados na area medica são analisados. / Mixture and non-mixture lifetime models are applied to analyze survival data when some individuals may never experience the event of interest. Dierent statistical models are proposed to analyze survival data in the presence of cure fraction. In this thesis, we propose the use of new models. From the univariate case, we consider that the lifetime data have a three-parameter Burr XII distribution, which includes the popular Weibull mixture model as a special case. We consider a general survival model where the scale and shape parameters of the Burr XII distribution depends on covariates. Also considering the univariate case the two-parameters exponentiated exponential distribution is used. The two-parameter exponentiated exponential or the generalized exponential distribution is a particular member of the exponentiated Weibull distribution introduced by Mudholkar and Srivastava (1993). We also consider in this case a general survival model where the scale, shape and cured fraction parameters of the exponentiated exponential distribution depends on covariates. We also introduce the univariate Weibull distributions in presence of cure fraction, censored data and covariates. Two models are explored in this case: the mixture model and non-mixture model. When we have two lifetimes associated with each unit (bivariate data), we can use some bivariate distributions: as special case the Block and Basu bivariate lifetime distribution. We also presents estimates for the parameters included in Block and Basu bivariate lifetime distribution in presence of covariates and cure fraction, applied to analyze survival data when some individuals may never experience the event of interest and two lifetimes are associated with each unit. We also consider in bivariate case the bivariate Weibull distributions derived from copula functions in presence of cure fraction, censored data and covariates. Two copula functions are explored in this paper: the Farlie-Gumbel-Morgenstern copula (FGM) and the Gumbel copula. Classical and Bayesian procedures are used to get point and condence intervals of the unknown parameters. Illustrations of the proposed methodologies are given considering medicals data sets.

Robust Flood Risk Analysis (FRA) is essential for effective flood risk management. The performance of any flood defence assets will heavily influence the estimate of an area's flood risk. It is therefore critical that the probability of a coastal flood defence asset incurring a structural failure when subjected to a particular loading i.e. its fragility is accurately quantified. The fragility representations of coastal defence assets presently adopted in UK National FRA (NaFRA) suffer three pertinent limitations. Firstly, assumptions relating to the modelling of the dependence structure of the variables that comprise the hydraulic load, including the water level, wave height and period, are restricted to a single loading variable. Consequently, due to the "system wide" nature of the analysis, a defence's conditional failure probability must also be expressed in terms of a single loading in the form of a fragility curve. For coastal defences the single loading is the overtopping discharge, an amalgamation of these basic loading variables. The prevalence of other failure initiation mechanisms may vary considerably for combinations of the basic loadings which give rise to equal overtopping discharges. Hence the univariate nature of the existing representations potentially restricts their ability to accurately assess an asset's structural vulnerability. Secondly, they only consider failure at least partially initiated through overtopping and thus neglect other pertinent initiation mechanisms acting in its absence. Thirdly, fragility representations have been derived for 61 generic assets (idealised forms of the defences found around the UK coast) each in five possible states of repair. The fragility representation associated with the generic asset and its state of repair deemed to most closely resemble a particular defence is adopted to describe its fragility. Any disparity in the parameters which influence the defence's structural vulnerability in the generic form of the asset and those observed in the field are also likely to further reduce the robustness of the existing fragility representations. In NaFRA coastal flood defence assets are broadly classified as vertical walls, beaches and embankments. The latter are typically found in sheltered locations where failure is water level driven and hence expressing failure probability conditionally on overtopping is admissible. Therefore new fragility representations for vertical wall and gravel beach assets which address the limitations of those presently adopted in NaFRA are derived. To achieve this aim it was necessary to propose new procedures for extracting information on the site and structural parameters characterising a defence's structural vulnerability from relevant resources (predominately beach profiles). In addition novel statistical approaches were put forward for capturing the uncertainties in the parameters on the basis of the site specific data obtained after implementation of the aforementioned procedures. A preliminary validation demonstrated the apparent reliability of these approaches. The pertinent initiation mechanisms behind the structural failure of each asset type were then identified before the state-of-the-art models for predicting the prevalence of these mechanisms during an event were evaluated. The Obhrai et al. (2008) re-formulation of the Bradbury (2000) barrier inertia model, which encapsulates all of the initiating mechanisms behind the structural failure of a beach, was reasoned as a more appropriate model for predicting the breach of a beach than that adopted in NaFRA. Failure initiated exclusively at the toe of a seawall was explicitly accounted for in the new formulations of the fragility representations using the predictors for sand and shingle beaches derived by Sutherland et al. (2007) and Powell & Lowe (1994). In order to assess whether the new formulations warrant a place in future FRAs they were derived for the relevant assets in Lyme Bay (UK). The inclusion of site specific information in the derivation of fragility representations resulted in a several orders of magnitude change in the Annual Failure Probabilities (AFPs) of the vertical wall assets. The assets deemed most vulnerable were amongst those assigned the lowest AFPs in the existing analysis. The site specific data indicated that the crest elevations assumed in NaFRA are reliable. Hence it appears the more accurate specification of asset geometry and in particular the inclusion of the beach elevation in the immediate vicinity of the structure in the overtopping calculation is responsible for the changes. The AFP was zero for many of the walls (≈ 77%) indicating other mechanism(s) occurring in the absence of any overtopping are likely to be responsible for failure. Toe scour was found to be the dominant failure mechanism at all of the assets at which it was considered a plausible cause of breach. Increases of at least an order of magnitude upon the AFP after the inclusion of site specific information in the fragility representations were observed at ≈ 86% of the walls. The AFPs assigned by the new site specific multivariate fragility representations to the beach assets were positively correlated with those prescribed by the existing representations. However, once the new representations were adopted there was substantially more variability in AFPs of the beach assets which had previously been deemed to be in identical states of repair. As part of the work, the new and existing fragility representations were validated at assets which had experienced failure or near-failure in the recent past, using the hydraulic loading conditions recorded during the event. No appraisal of the reliability of the new representations for beaches was possible due to an absence of any such events within Lyme Bay. Their AFPs suggest that armed with more information about an asset's geometry the new formulations are able to provide a more robust description of a beach's structural vulnerability. The results of the validation as well as the magnitude of the AFPs assigned by the new representations on the basis of field data suggest that the newly proposed representations provide the more realistic description of the structural vulnerability of seawalls. Any final conclusions regarding the robustness of the representations must be deferred until more failure data becomes available. The trade-off for the potentially more robust description of an asset's structural vulnerability was a substantial increase in the time required for the newly derived fragility representations to compute the failure probability associated with a hydraulic loading event. To combat this increase, (multivariate) generic versions of the new representations were derived using the structural specific data from the assets within Lyme Bay. Although there was generally good agreement in the failure probabilities assigned to the individual hydraulic loading events by the new generic representations there was evidence of systematic error. This error has the potential to bias flood risk estimates and thus requires investigation before the new generic representations are included in future FRAs. Given the disparity in the estimated structural vulnerability of the assets according to the existing fragility curves and the site-specific multivariate representations the new generic representations are likely to be more reliable than the existing fragility curves.