- Prof. Thomas Mikosch – University of Copenhagen
- FCUL – DEIO – Bloco C/2 – Piso 2 – Sala 8.2.03 – 14:30
- Sexta-feira, 15 de Novembro de 2002
In this lecture, we consider some of the standard models of financial time series and study the interplay between the dependence and the tail behavior. This relationship is particularly important if one wants to study the extremal behavior of a time series. Starting with some evidence from return data, we check whether these models do what they are supposed to. In particular, we consider a flexible class of multivariate heavy-tailed distributions, the regularly varying distributions. It turns out that various standard models, such as GARCH, have regularly varying finite-dimensional distributions. We also stress that correlations and covariances are not appropriate measures of dependence for heavy-tailed time series. This lecture is based on some notes available under: www.math.ku.dk/mikosch/semstat.