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
Dag Tjostheim, Bergen U. Print
Thursday, 06 May 2010, 12:15 - 13:15

Gaussian local correlation as a measure of nonlinear dependence

Dag Tjostheim, University of Bergen

Abstract: The most commonly used dependence measure for random variables is the correlation; in the time series case the autocorrelation function. It is very easy to find nonlinear examples where the correlation completely fails to describe the dependence involved. The concept of correlation is primarily useful for Gaussian variables. We introduce a new measure of local dependence by locally approximating the joint density of two (or more) random variables by a family of Gaussian densities and by using the correlation of the approximating Gaussian as a local correlation. Connections to tail dependence and measuring skewness of the dependence pattern of finance data will be mentioned, and the usefulness of this concept will be sought demonstrated in a  number of different contexts for real and simulated data.

Location: S 12.227
Contact: Claude Adan, This e-mail address is being protected from spam bots, you need JavaScript enabled to view it