September 1, 2006
Recently, there has been increasing interest in forecasting methods that utilise large datasets. We explore the possibility of forecasting with model averaging using the out-of-sample forecasting performance of various models in a frequentist setting, using the predictive likelihood. We apply our method to forecasting UK inflation and find that the new method performs well; in some respects it outperforms other averaging methods.
J.E.L classification codes: C110, C150, C530
Keywords:Forecasting, Inflation, Bayesian model averaging, Akaike criterion, Forecast combining