Group for Research in Decision Analysis

Computation and Inference for Stochastic Volatility Models Driven by Levy Processes

David Stephens

The standard models used in finanical time series applications are often deficient in that they do not capture observed behaviour adequately. Specifically, the standard models do not capture the time-varying or stochastic nature of the process variance. I will describe some continuous time models, constructed via stochastic differential equations driven by Levy processes - stochastically continuous random processes with independent increments - that better capture observed characteristics of financial time series data.

I will illustrate the implementation of the models on several real data sets, and identify some of the computational challenges.