Menu Content/Inhalt
Seminars Print
previous year previous month next month next year
See by year See by month See by week See Today Search Jump to month
Olivier Wintenberger, Paris Dauphine Print
Friday, 07 October 2011, 14:30 - 15:30

Olivier Wintenberger, Paris Dauphine

Detecting multiple change points using Quasi Likelihood

Abstract:
This talk is devoted to the off-line multiple change-point detection using the Quasi Likelihood criteria. The abstract theoretical framework is large in order to consider AR(∞), ARCH(∞), TARCH(∞),... models in one blow. A penalization is proposed in order to infer the unknown number of change points robustly with respect to the dependence of the observations.  The estimation procedure of the change points and the parameters of the model is consistent under second order moments and asymptotically normal under fourth order moments. In practice the penalization is calibrated via a new slope heuristic and the inference is done in a reasonable ($O(n^2)$) time. Simulations and an application on real data are given.

Location: Plaine, NO 9th floor - salle des profs
Contact: This e-mail address is being protected from spam bots, you need JavaScript enabled to view it

Back