School of Mathematics
Cardiff University
(Statistics and Operational Research Seminar)
ON THE CONTROVERSIES OVER TAILWEIGHT AND LONG RANGE DEPENDENCE
C. C. HEYDE
Australian National University and Columbia University
*The behaviour of most probabilistic models is critically influenced by the tails of the
distributions which drive the models. This is especially important in areas such as asset
pricing, reliability, insurance risk, and telecommunications networks. Nevertheless, there
remains much uncertainty, controversy and error in connection with distinctions between
tail weights, even those between power and exponential tails. In this talk it will be
explained why the distinctions are surprisingly difficult, why particularly large samples are
necessary for clear discrimination, and why theoretical errors have confused the picture.
*Although there has been a burgeoning literature related to the existence of long range
dependence (LRD) in the telegraphic and finance areas, there has been a lag with
applications which capitalize on these insights. This is partly a result of the limited intuitive
content and inflexibility in the definition(s) which basically focus on covariance properties.
In this talk the theory will be taken away from its covariance setting and a focus given to
modes of LRD, namely persistence of signs and/or persistence of magnitudes, beyond
what can be expected under weak dependence. An application to discrimination between
risky asset pricing models will be given to illustrate the ideas.
28 November 2002 Thursday 16.00
Room M/0.37, School Of Mathematics, Cardiff University, Senghennyd Road, Cardiff
|