This thesis consists of three chapters. The first chapter focuses on adaptive smoothing splines for fitting functions with varying roughness. In the first part of the first chapter, we study an asymptotically optimal procedure to choose the value of a discretized version of the variable smoothing parameter in adaptive smoothing splines. With the choice given by the multivariate version of the generalized cross validation, the resulting adaptive smoothing spline estimator is shown to be consistent and asymptotically optimal under some general conditions. In the second part, we derive the asymptotically optimal local penalty function, which is subsequently used for the derivation of the locally optimal smoothing spline estimator. In the second chapter, we propose a Lipschitz regularity based statistical model, and apply it to coordinate measuring machine (CMM) data to estimate the form error of a manufactured product and to determine the optimal sampling positions of CMM measurements. Our proposed wavelet-based model takes advantage of the fact that the Lipschitz regularity holds for the CMM data. The third chapter focuses on the classification of functional data which are known to be well separable within a particular interval. We propose an interval based classifier. We first estimate a baseline of each class via convex optimization, and then identify an optimal interval that maximizes the difference among the baselines. Our interval based classifier is constructed based on the identified optimal interval. The derived classifier can be implemented via a low-order-of-complexity algorithm.

This thesis proposes some new model independent or nonparametric methods for estimating the dose-response curve and the effective dosage curve in the context of bioassay. The research problem is also of importance in environmental risk assessment and other areas of health sciences. It is shown in the thesis that our new nonparametric methods while bearing optimal asymptotic properties also exhibit strong finite sample performance. Although our specific emphasis is on bioassay and environmental risk assessment, the methodology developed in this dissertation applies broadly to general order restricted inference.

We propose a new method to perform functional principal component analysis (FPCA) for discretely observed functional data by solving successive optimization problems. The new framework can be applied to both regularly and irregularly observed data, and to both dense and sparse data. Our method does not require estimates of the individual sample functions or the covariance functions. Hence, it can be used to analyze functional data with multidimensional arguments (e.g. random surfaces). Furthermore, it can be applied to many processes and models with complicated or nonsmooth covariance functions. In our method, smoothness of eigenfunctions is controlled by directly imposing roughness penalties on eigenfunctions, which makes it more efficient and flexible to tune the smoothness. Efficient algorithms for solving the successive optimization problems are proposed. We provide the existence and characterization of the solutions to the successive optimization problems. The consistency of our method is also proved. Through simulations, we demonstrate that our method performs well in the cases with smooth samples curves, with discontinuous sample curves and nonsmooth covariance and with sample functions having two dimensional arguments (random surfaces), repectively. We apply our method to classification problems of retinal pigment epithelial cells in eyes of mice and to longitudinal CD4 counts data. In the second part of this dissertation, we propose a sparse Fisher’s discriminant analysis method with thresholded linear constraints. Various regularized linear discriminant analysis (LDA) methods have been proposed to address the problems of the LDA in high-dimensional settings. Asymptotic optimality has been established for some of these methods when there are only two classes. A difficulty in the asymptotic study for the multiclass classification is that for the two-class classification, the classification boundary is a hyperplane and an explicit formula for the classification error exists, however, in the case of multiclass, the boundary is usually complicated and no explicit formula for the error generally exists. Another difficulty in proving the asymptotic consistency and optimality for sparse Fisher’s discriminant analysis is that the covariance matrix is involved in the constraints of the optimization problems for high order components. It is not easy to estimate a general high-dimensional covariance matrix. Thus, we propose a sparse Fisher’s discriminant analysis method which avoids the estimation of the covariance matrix, provide asymptotic consistency results and the corresponding convergence rates for all components. To prove the asymptotic optimality, we provide an asymptotic upper bound for a general linear classification rule in the case of muticlass which is applied to our method to obtain the asymptotic optimality and the corresponding convergence rate. In the special case of two classes, our method achieves the same as or better convergence rates compared to the existing method. The proposed method is applied to multivariate functional data with wavelet transformations.

The dissertation investigates asymptotic theory of decentralized sequential hypothesis testing problems as well as asymptotic behaviors of the Sequential Minimum Energy Design (SMED). The main results are summarized as follows. 1.We develop the first-order asymptotic optimality theory for decentralized sequential multi-hypothesis testing under a Bayes framework. Asymptotically optimal tests are obtained from the class of "two-stage" procedures and the optimal local quantizers are shown to be the "maximin" quantizers that are characterized as a randomization of at most M-1 Unambiguous Likelihood Quantizers (ULQ) when testing M >= 2 hypotheses. 2. We generalize the classical Kullback-Leibler inequality to investigate the quantization effects on the second-order and other general-order moments of log-likelihood ratios. It is shown that a quantization may increase these quantities, but such an increase is bounded by a universal constant that depends on the order of the moment. This result provides a simpler sufficient condition for asymptotic theory of decentralized sequential detection. 3. We propose a class of multi-stage tests for decentralized sequential multi-hypothesis testing problems, and show that with suitably chosen thresholds at different stages, it can hold the second-order asymptotic optimality properties when the hypotheses testing problem is "asymmetric." 4. We characterize the asymptotic behaviors of SMED algorithm, particularly the denseness and distributions of the design points. In addition, we propose a simplified version of SMED that is computationally more efficient.

Cette thèse est constituée de deux parties qui peuvent être lues indépendamment. Dans la première partie de la thèse, nous étudions des problèmes de couverture et de valorisation d’options liés à une mesure de risque. Notre approche principale est l’utilisation d’une fonction de risque asymétrique et d’un cadre asymptotique dans lequel nous obtenons des solutions optimales à travers des équations aux dérivées partielles (EDP) non-linéaires.Dans le premier chapitre, nous nous intéressons à la valorisation et la couverture des options européennes. Nous considérons le problème de l’optimisation du risque résiduel généré par une couverture à temps discret en présence d’un critère asymétrique de risque. Au lieu d'analyser le comportement asymptotique de la solution du problème discret associé, nous avons étudié la mesure asymétrique du risque résiduel intégré dans un cadre Markovian. Dans ce contexte, nous montrons l’existence de cette mesure de risque asymptotique. Ainsi, nous décrivons une stratégie de couverture asymptotiquement optimale via la solution d’une EDP totalement non-linéaire.Le deuxième chapitre est une application de cette méthode de couverture au problème de valorisation de la production d’une centrale. Puisque la centrale génère de coûts de maintenance qu’elle soit allumée ou non, nous nous sommes intéressés à la réduction du risque associé aux revenus incertains de cette centrale en se couvrant avec des contrats à terme. Nous avons étudié l’impact d’un coût de maintenance dépendant du prix d’électricité dans la stratégie couverture.Dans la seconde partie de la thèse, nous considérons plusieurs problèmes de contrôle liés à l'économie et la finance.Le troisième chapitre est dédié à l’étude d’une classe de problème du type McKean-Vlasov (MKV) avec bruit commun, appelée MKV polynomiale conditionnelle. Nous réduisons cette classe polynomiale par plongement de Markov à des problèmes de contrôle en dimension finie.Nous comparons trois techniques probabilistes différentes pour la résolution numérique du problème réduit: la quantification, la régression par randomisation du contrôle et la régression différée. Nous fournissons de nombreux exemples numériques, comme par exemple, la sélection de portefeuille avec incertitude sur une tendance du sous-jacent.Dans le quatrième chapitre, nous résolvons des équations de programmation dynamique associées à des valorisations financières sur le marché de l’énergie. Nous considérons qu’un modèle calibré pour les sous-jacents n’est pas disponible et qu’un petit échantillon obtenu des données historiques est accessible.En plus, dans ce contexte, nous supposons que les contrats à terme sont souvent gouvernés par des facteurs cachés modélisés par des processus de Markov. Nous proposons une méthode nonintrusive pour résoudre ces équations à travers les techniques de régression empirique en utilisant seulement l’historique du log du prix des contrats à terme observables. / This thesis is constituted by two parts that can be read independently.In the first part, we study several problems of hedging and pricing of options related to a risk measure. Our main approach is the use of an asymmetric risk function and an asymptotic framework in which we obtain optimal solutions through nonlinear partial differential equations (PDE).In the first chapter, we focus on pricing and hedging European options. We consider the optimization problem of the residual risk generated by a discrete-time hedging in the presence of an asymmetric risk criterion. Instead of analyzing the asymptotic behavior of the solution to the associated discrete problem, we study the integrated asymmetric measure of the residual risk in a Markovian framework. In this context, we show the existence of the asymptotic risk measure. Thus, we describe an asymptotically optimal hedging strategy via the solution to a fully nonlinear PDE.The second chapter is an application of the hedging method to the valuation problem of the power plant. Since the power plant generates maintenance costs whether it is on or off, we are interested in reducing the risk associated with its uncertain revenues by hedging with forwards contracts. We study the impact of a maintenance cost depending on the electricity price into the hedging strategy.In the second part, we consider several control problems associated with economy and finance.The third chapter is dedicated to the study of a McKean-Vlasov (MKV) problem class with common noise, called polynomial conditional MKV. We reduce this polynomial class by a Markov embedding to finite-dimensional control problems.We compare three different probabilistic techniques for numerical resolution of the reduced problem: quantization, control randomization and regress later.We provide numerous numerical examples, such as the selection of a portfolio under drift uncertainty.In the fourth chapter, we solve dynamic programming equations associated with financial valuations in the energy market. We consider that a calibrated underlying model is not available and that a limited sample of historical data is accessible.In this context, we suppose that forward contracts are governed by hidden factors modeled by Markov processes. We propose a non-intrusive method to solve these equations through empirical regression techniques using only the log price history of observable futures contracts.