A0535
Title: Stable quasi-likelihood regression
Authors: Hiroki Masuda - University of Tokyo (Japan) [presenting]
Abstract: Some parametric estimation results are presented for a regression model that can cover not only stochastic differential equations but also some semimartingale regression models. We introduce a class of non-Gaussian quasi-likelihood estimators, and prove its asymptotic mixed normality under the locally stable property of the driving noise process. In sharp contrast to the Gaussian quasi-likelihood based counterpart, the result holds without any moment condition and any stability even when attempting to estimate the trend parameter. Some numerical experiments are shown to illustrate effectiveness of the proposed methodology.
