This thesis is dedicated to the development of statistical and computational methods for the analysis of DNA sequences and gene expression time series.<br /><br />First we study a parsimonious Markov model called Mixture Transition Distribution (MTD) model which is a mixture of Markovian transitions. The overly high number of constraints on the parameters of this model hampers the formulation of an analytical expression of the Maximum Likelihood Estimate (MLE). We propose to approach the MLE thanks to an EM algorithm. After comparing the performance of this algorithm to results from the litterature, we use it to evaluate the relevance of MTD modeling for bacteria DNA coding sequences in comparison with standard Markovian modeling.<br /><br />Then we propose two different approaches for genetic regulation network recovering. We model those genetic networks with Dynamic Bayesian Networks (DBNs) whose edges describe the dependency relationships between time-delayed genes expression. The aim is to estimate the topology of this graph despite the overly low number of repeated measurements compared with the number of observed genes. <br /><br />To face this problem of dimension, we first assume that the dependency relationships are homogeneous, that is the graph topology is constant across time. Then we propose to approximate this graph by considering partial order dependencies. The concept of partial order dependence graphs, already introduced for static and non directed graphs, is adapted and characterized for DBNs using the theory of graphical models. From these results, we develop a deterministic procedure for DBNs inference. <br /><br />Finally, we relax the homogeneity assumption by considering the succession of several homogeneous phases. We consider a multiple changepoint<br />regression model. Each changepoint indicates a change in the regression model parameters, which corresponds to the way an expression level depends on the others. Using reversible jump MCMC methods, we develop a stochastic algorithm which allows to simultaneously infer the changepoints location and the structure of the network within the phases delimited by the changepoints. <br /><br />Validation of those two approaches is carried out on both simulated and real data analysis.

Dans cette thèse, je me suis interessé a deux aspects de la gestion de portefeuille : la maximisation de l'utilité e d'un portefeuille financier lorsque on impose une contrainte sur l'exposition au risque, et la couverture quadratique en marché incomplet. Part I. Dans la première partie, j' étudie un problème d'assurance de portefeuille du point de vue du manager d'un fond d'investissement, qui veut structurer un produit financier pour les investisseurs du fond avec une garantie sur la valeur du portefeuille a la maturité . Si, a la maturité, la valeur du portefeuille est au dessous d'un seuil x e, l'investisseur sera remboursé a la hauteur de ce seuil par une troisième partie, qui joue le rôle d'assureur du fond (on peut imaginer que le fond appartient à une banque et que donc c'est la banque elle même qui joue le rôle d'assureur). En échange de cette assurance, la troisième partie impose une contrainte sur l'exposition au risque que le manager du fond peut tolérer, mesurée avec une mesure de risque monétaire convexe. Je donne la solution complet e de ce problème de maximisation non convexe en marché complet et je prouve que le choix de la mesure de risque est un point crucial pour avoir existence d'un portefeuille optimal. J'applique donc mes résultats lorsque on utilise la mesure de risque entropique (pour laquelle le portefeuille optimal existe toujours), les mesures de risque spectrales (pour lesquelles le portefeuille optimal peut ne pas exister dans certains cas) et la G-divergence. Mots-cl es : Assurance de portefeuille ; maximisation d'utilité ; mesure de risque convexe ; VaR, CVaR et mesure de risque spectrale ; entropie et G-divergence. Part II. Dans la deuxième partie, je m'intéresse au problème de couverture quadratique en marché incomplet. J'assume que le marché est d écrit par un processus Markovien tridimensionnel avec sauts. La premi ère variable d' état décrit l'actif - financier, échangeable sur le marché, qui sert comme instrument de couverture ; la deuxième variable d' état modélise un actif financier que intervient dans la dynamique de l'instrument de couverture mais qui n'est pas échangeable sur le march é : il peut donc être vu comme un facteur de volatilité de l'instrument de couverture, ou comme un actif financier que l'on ne peut pas acheter (pour de raisons légales par exemple) ; la troisième et dernière variable d' état représente une source externe de risque qui affecte l'option Européenne qu'on veut couvrir, et qui, elle aussi, n'est pas échangeable sur le marché. Pour résoudre le problème j'utilise l'approche de la programmation dynamique, qui me permet d' écrire l' équation de Hamilton-Jacobi- Bellman associé e au problème de couverture quadratique, qui est non locale en non linéaire. Je prouve que la fonction valeur associée au problème de couverture quadratique peut être caractérisée par un système de trois équations integro- différentielles aux dérivées partielles, dont l'une est semilinéaire et ne dépends pas du choix de l'option a couvrir, et les deux autres sont simplement linéaires , et que ce système a une unique solution r régulière dans un espace de Hölder approprié, qui me permet donc de caractériser la stratégie de couverture optimale . Ce résultat est démontré lorsque le processus est non dégénéré (c'est a dire que la composante Brownienne est strictement elliptique) et lorsque le processus est a sauts purs. Je conclus avec une application de mes résultats dans le cadre du marché de l' électricité. Mots-cl es : Couverture quadratique ; modèle a sauts ; programmation dynamique ; équation de Hamilton-Jacobi-Bellman ; équations aux dérivées partielles integro-différentielles.

<p>In this thesis the double exponential jump-diffusion model is considered and the Laplace transform is used as a method for pricing both plain vanilla and path-dependent options. The evolution of the underlying stock prices are assumed to follow a double exponential jump-diffusion model. To invert the Laplace transform, the Euler algorithm is used. The thesis includes the programme code for European options and the application to the real data. The results show how the Kou model performs on the NASDAQ OMX Stockholm Market in the case of the SEB stock.</p>

In this thesis the double exponential jump-diffusion model is considered and the Laplace transform is used as a method for pricing both plain vanilla and path-dependent options. The evolution of the underlying stock prices are assumed to follow a double exponential jump-diffusion model. To invert the Laplace transform, the Euler algorithm is used. The thesis includes the programme code for European options and the application to the real data. The results show how the Kou model performs on the NASDAQ OMX Stockholm Market in the case of the SEB stock.

在迴歸分析中，若變數間具有非線性的關係時，B-Spline線性迴歸是以無母數的方式建立模型。B-Spline函數為具有節點(knots)的分段多項式，選取合適節點的位置對B-Spline的估計有重要的影響，在近年來許多的文獻中已提出一些尋找節點位置的估計方法，而本文中我們提出了一種基於Bayesian information criterion(BIC)的節點估計方式。 我們想要深入了解在不同類型的迴歸函數間，各種選取節點方法的配適效果與模擬時間，並且加以比較，在使用B-Spline函數估計時，能夠使用合適的方法尋找節點。 / In regression analysis, when the relation between the response variable and the explanatory variable is nonlinear, one can use nonparametric methods to estimate the regression function. B-Spline regression is one of the popular nonparametric regression methods. B-Splines are piecewise polynomial joint at knots, and the choice of knot locations is crucial. Zhou and Shen (2001) proposed to use spatially adaptive regression splines (SARS), where the knots are estimated using a selection scheme. Dimatteo, Genovese, and Kass (2001) proposed to use Bayesian adaptive regression splines (BARS), where certain priors for knot locations are considered. In this thesis, a knot estimation method based on the Bayesian information criterion (BIC) is proposed, and simulation studies are carried out to compare BARS, SARS and the proposed BIC-based method.

By allowing to measure the magnetic field distribution inside a material, muon spin rotation experiments have the potential to provide valuable information about microscopic properties of high-temperature superconductors. Nevertheless, information about the intrinsic superconducting properties of the material is masked by random thermal and static fluctuations of the magnetic field which penetrates the material in the form of vortices of quantized magnetic flux. A good understanding of the fluctuations of those vortices is needed for the correct determination of intrinsic properties, notably the coherence length &xi;, and the field penetration depth &lambda;. We develop a simulation based on the Metropolis algorithm in order to understand the effect, on the magnetic field distribution, of disorder- and thermally-induced fluctuations of the vortex lattice inside a layered superconductor. <br /><br /> Our model correctly predicts the melting temperatures of the YBa₂Cu₃O_{6. 95} (YBCO) superconductor but largely underestimates the observed entropy jump. Also we failed to simulate the high field disordered phase, possibly because of a finite size limitation. In addition, we found our model unable to describe the first-order transition observed in the highly anisotropic Bi₂Sr₂CaCu₂O_8+<em>y</em>. <br /><br /> Our model predicts that for YBCO, the effect of thermal fluctuations on the field distribution is indistinguishable from a change in &xi;. It also confirms the usual assumption that the effect of static fluctuations at low temperature can be efficiently modeled by convolution of the field distribution with a Gaussian function. However the extraction of &xi; at low fields requires a very high resolution of the field distribution because of the low vortex density.

By allowing to measure the magnetic field distribution inside a material, muon spin rotation experiments have the potential to provide valuable information about microscopic properties of high-temperature superconductors. Nevertheless, information about the intrinsic superconducting properties of the material is masked by random thermal and static fluctuations of the magnetic field which penetrates the material in the form of vortices of quantized magnetic flux. A good understanding of the fluctuations of those vortices is needed for the correct determination of intrinsic properties, notably the coherence length &xi;, and the field penetration depth &lambda;. We develop a simulation based on the Metropolis algorithm in order to understand the effect, on the magnetic field distribution, of disorder- and thermally-induced fluctuations of the vortex lattice inside a layered superconductor. <br /><br /> Our model correctly predicts the melting temperatures of the YBa₂Cu₃O_{6. 95} (YBCO) superconductor but largely underestimates the observed entropy jump. Also we failed to simulate the high field disordered phase, possibly because of a finite size limitation. In addition, we found our model unable to describe the first-order transition observed in the highly anisotropic Bi₂Sr₂CaCu₂O_8+<em>y</em>. <br /><br /> Our model predicts that for YBCO, the effect of thermal fluctuations on the field distribution is indistinguishable from a change in &xi;. It also confirms the usual assumption that the effect of static fluctuations at low temperature can be efficiently modeled by convolution of the field distribution with a Gaussian function. However the extraction of &xi; at low fields requires a very high resolution of the field distribution because of the low vortex density.

We propose multigrid methods for solving Hamilton-Jacobi-Bellman (HJB) and Hamilton-Jacobi-Bellman-Isaacs (HJBI) equations. The methods are based on the full approximation scheme. We propose a damped-relaxation method as smoother for multigrid. In contrast with policy iteration, the relaxation scheme is convergent for both HJB and HJBI equations. We show by local Fourier analysis that the damped-relaxation smoother effectively reduces high frequency error. For problems where the control has jumps, restriction and interpolation methods are devised to capture the jump on the coarse grid as well as during coarse grid correction. We will demonstrate the effectiveness of the proposed multigrid methods for solving HJB and HJBI equations arising from option pricing as well as problems where policy iteration does not converge or converges slowly.

Guaranteed Minimum Withdrawal Benefits(GMWB) have become popular riders on variable annuities. The pricing of a GMWB contract was originally formulated as a singular stochastic control problem which results in a Hamilton Jacobi Bellman (HJB) Variational Inequality (VI). A penalty method method can then be used to solve the HJB VI. We present a rigorous proof of convergence of the penalty method to the viscosity solution of the HJB VI assuming the underlying asset follows a Geometric Brownian Motion. A direct control method is an alternative formulation for the HJB VI. We also extend the HJB VI to the case of where the underlying asset follows a Poisson jump diffusion. The HJB VI is normally solved numerically by an implicit method, which gives rise to highly nonlinear discretized algebraic equations. The classic policy iteration approach works well for the Geometric Brownian Motion case. However it is not efficient in some circumstances such as when the underlying asset follows a Poisson jump diffusion process. We develop a combined fixed point policy iteration scheme which significantly increases the efficiency of solving the discretized equations. Sufficient conditions to ensure the convergence of the combined fixed point policy iteration scheme are derived both for the penalty method and direct control method. The GMWB formulated as a singular control problem has a special structure which results in a block matrix fixed point policy iteration converging about one order of magnitude faster than a full matrix fixed point policy iteration. Sufficient conditions for convergence of the block matrix fixed point policy iteration are derived. Estimates for bounds on the penalty parameter (penalty method) and scaling parameter (direct control method) are obtained so that convergence of the iteration can be expected in the presence of round-off error.

A robust computational framework is presented for the risk-neutral valuation of capital guarantees written on discretely-reallocated portfolios following the Constant Proportion Portfolio Insurance (CPPI) strategy. Aiming to address the (arguably more realistic) cases where analytical results are unavailable, this framework accommodates risky-asset jumps, volatility surfaces, borrowing restrictions, nonuniform reallocation schedules and autonomous CPPI floor trajectories. The two-asset state space representation developed herein facilitates visualising the CPPI strategy, which in turn provides insight into grid design and interpolation. It is demonstrated that given a deterministic process for the risk-free rate, the pricing problem can be cast as solving cascading systems of 1D partial integro-differential equations (PIDEs). This formulation’s stability and monotonicity are studied. In addition to making more sense financially, the limited borrowing variant of the CPPI strategy is found to be better suited than the classical (unlimited borrowing) counterpart for bounded-domain calculations. Consequently, it is demonstrated how the unlimited borrowing problem can be approximated by imposing an artificial borrowing limit. For implementation validation, analytical solutions to special cases are derived. Numerical tests are presented to demonstrate the versatility of this framework.