Title: Bayesian Lasso for large vector autoregression models
Authors: Deborah Gefang - University of Leicester (United Kingdom) [presenting]
Abstract: Recent years have witnessed a growing interest in using Bayesian vector autoregression model (BVAR) that consists of many variables to gain deeper insight from large macroeconomic and financial data sets. Most of the available Bayesian techniques for large BVAR models, however, are subject to two types of constraints. The first type arises from the use of subjective priors which might not always be in line with the data generating process. The second type of constraint is associated with the computing power of the available computers, including the high performance computers. A Bayesian LASSO VAR method is developed to tackle these problems. Using FRED-MD monthly data for Macroeconomic Research provided by McCracken, we provide a comprehensive comparison between the forecasting performance of our method and that of other popular estimation techniques.