A0154
Title: The Bayesian lasso for variable selection of realized measures in a realized EGARCH model
Authors: Richard Gerlach - University of Sydney (Australia) [presenting]
Vica Tendenan - The University of Sydney (Australia)
Chao Wang - The University of Sydney (Australia)
Abstract: The Realized EGARCH specification includes multiple realized measures in a financial time series volatility model. However, the question remains about which and how many realized measures to include in the model. The purpose is to employ the Lasso method in a Bayesian context to assist in selecting among a range of realized measures for inclusion. Markov chain Monte Carlo methods are designed for this purpose. We investigate the impacts of our approach on parameter estimation and forecasting of volatility and tail risk. Several competing models are employed for forecast comparison.
