Simulation-based inference and nonlinear canonical analysis in financial econometrics.

摘要

The objective of this thesis is to study standard and simulation-based inference techniques which are valid in finite samples for models used in finance.; In the first essay, we study a simple moment estimator, available in closed form for general regression models with stochastic volatility models This easy-to-use estimator allows for simulation-based inference techniques which can be computationally expensive.; In the second essay, we exploit the closed-form expression of the moment estimator for the parameters of the SV model to implement simulation-based inference techniques such as Monte Carlo (MC) tests [see Dwass (1957), Barnard (1963), Birnbaum (1974)].; In the third essay, we estimate the SV model by indirect inference [see Smith (1993), Gourieroux, Monfort and Renault (1993), henceforth (GMR)] under nonregular conditions.; In the fourth essay, we characterize the one-dimensional stochastic differential equations, for which the eigenfunctions of the infinitesimal generator are polynomials in y. In particular, affine transformations of the Ornstein-Uhlenbeck process, the Cox-Ingersoll-Ross process and the Jacobi process belong to this stochastic differential equations family. Such processes exhibit specific patterns of the drift and volatility functions together with a particular form of the eigenvalues.; In the fifth essay, we consider a discretely sampled Jacobi process appropriate to specify the dynamics of a process with range [0,1], such as a discount coefficient, a regime probability, or a state price. (Abstract shortened by UMI.)