A0523
Title: Index-augmented autoregressive models: Representation, estimation, and forecasting
Authors: Elisa Scambelloni - University of Rome Tor Vergata (Italy)
Gianluca Cubadda - University of Rome Tor Vergata (Italy) [presenting]
Abstract: The purpose is to examine the condition under which each individual series that is generated by a $n-$dimensional Vector Auto-Regressive (VAR) model can be represented as an autoregressive model that is augmented with the lags of $q$ linear combinations of all the variables in the system. We call this modelling Index-Augmented Autoregression (IAAR). We show that the parameters of the IAAR can be estimated by a switching algorithm that increases the Gaussian likelihood at each iteration. Provided that the number of factors $q$ times the VAR order $p$ is small compared to the sample size $T$, we propose a regularized version of our algorithm to handle a medium-large number of time series. We illustrate the usefulness of the IAAR modelling both by empirical applications and simulations.
