Time: Monday, 15. March 2004 from 13.00 to 14.30 Location: Seminarraum 105a (Mehrzweckraum), Argentinierstr. 8, 1. Floor
Speaker: Dr. Daniel Straumann (RiskLab, ETH Zurich)
Title: Estimation in Conditionally Heteroscedastic Time Series Models
Abstract: This talk deals with the estimation in certain conditionally heteroscedastic time series models, such as e.g. GARCH, asymmetric GARCH or EGARCH. By exploiting the techniques of stochastic recurrence equations, we develop a general and unifying limit theory for the maximum-likelihood estimator (MLE) and quasi-maximum likelihood estimator (QMLE) in a certain parametric class of conditionally heteroscedastic time series models. This generalizes and clarifies work of Lumsdaine (1996) and Berkes et al.(2003). We furthermore discuss the issue of misspecification in the MLE and the behavior of the QMLE in the presence of a heavy-tailed noise distribution. A second part of the thesis studies the asymptotic behavior of the classical Whittle estimator when it is applied to the squares of GARCH(1,1). We focus on the case of an unconditional distribution with an infinite 8th moment and thereby generalize results by Giraitis and Robinson (2001).
A copy of Dr. Straumann's recent Ph.D. thesis on this topic is in my office.
With best regards,
Uwe Schmock