Fast Methods for High-Dimensional Financial Time Series Modeling

• Activity Number: 160
• Type: Topic Contributed
• Date/Time: Monday, July 30, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Business and Economic Statistics Section
• Author(s): Nalini Ravishanker*+ and Jeffrey S. Pai
• Companies: University of Connecticut and University of Manitoba
• Keywords: Fast algorithms ; Financial time series ; Bayesian

Abstract

This talk describes methods for fast modeling and prediction of long multiple time series. Such series arise in several disciplines including finance, and often exhibit features such as nonlinearity, persistence, and volatility. We describe a computationally feasible approach for likelihood based estimation. The method is based on a multivariate preconditioned conjugate gradient (MPCG) algorithm, involving solution of a block-Toeplitz system, and Monte Carlo integration over unobserved latent variables. We illustrate our approach for pricing financial derivatives related to weather, as well as stock returns series. Part of this research is joint work with Jeffrey Pai, University of Manitoba.