A1118
Title: Forecasting with and maximum likelihood estimation of the vector autoregressive to anything (VARTA) model
Authors: Jonas Andersson - Norwegian School of Economics (Norway) [presenting]
Dimitris Karlis - RC Athens University of Economics and Business (Greece)
Abstract: The literature on multivariate time series is largely limited to either models based on the multivariate Gaussian distribution or models specifically developed for a given application. A general approach is developed based on an underlying, unobserved Gaussian Vector Autoregressive (VAR) model. Using a transformation, the time dynamics, as well as the distributional properties of a multivariate time series, can be captured. The model is called the Vector AutoRegressive to Anything (VARTA) model and was originally presented in a prior study, where it was used for the purpose of simulation. A maximum likelihood estimator is derived for the model and its performance is investigated. Diagnostic analysis and methods for predictive distribution computation are provided. The modelling approach is applied to a multivariate time series of wind speeds in nearby locations.