Annual Report 2008.pdf - SAMSI

one of the main sources of uncertainty. Additionally, fundamental seasonal demand and supply
factors are known to influence price levels.
In the present, accurate VaR forecasts are obtained applying a two steps procedure. In the first
step, both the conditional mean and variance of price series are modeled. Next, the gotten
residuals are treated using extreme values methodology. In this work the electricity price series
for the Spanish Market are modeled using the aforementioned procedure. Specifically, an
ARIMA model is used for the mean equation. Traditionally, the conditional variance is estimated
using models from the GARCH family. In this work, two models from the GARCH family are
used, but also a stochastic volatility model is considered. Comparative study results are shown.
John Nolan
American University
jpnolan@american.edu
“Stable Laws and Extreme Value Laws”
Stable distributions are a class of heavy tailed distributions that generalize the Gaussian
distribution and that can be used to model in a variety of problems. An overview of univariate
stable laws is given, with emphasis on the practical aspects of working with stable distributions.
Connections between stable laws and extreme value laws will be discussed. Then multivariate
stable distributions and multivariate extreme value laws will be examined, with some
connections explored.
Luis Pericchi
Universidad de Puerto Rico, San Juan and Universidad Simon Bolivar, Caracas
luarpr@gmail.com
“Experiences with Extreme Data in the Caribbean: Bayes, Re-Parametrizations and the
Multivariate Approach of Heffernan and Tawn”
We review several statistical analyses of univariate data with seasonalities and clustering, as well
as multivariate data, with data from Venezuela, Central America and Puerto Rico. Emphasis is
placed on: i)The importance of the Bayesian Approach to keep all sources of uncertainty, ii) The
convenience of orthogonal re-parameterizations the Generalized Extreme Value Distribution and
Pareto, and iii) The usefulness and open issues of the Heffernan and Tawn (2004) approach.
These are joint works with Stuart Coles (University of Padova), Scott Sisson (University of New
South Wales) and Beatriz Mendes (Federal University of Rio de Janeiro).
Sidney Resnick
Cornell University
sir1@cornell.edu
“Conditioned Limit Theorems: Does the Story End with a Bang or a Whimper?”