That is, it is not valid to look at how well a model fits the historical data. The accuracy of forecasts can only be determined by considering how well a model performs on data that were not used when fitting the model [15]. When comparing different models, it is common to use a portion of the available data for fitting – the in-sample data, and use the rest of the data to measure how well the model is likely to forecast on new data – the out-of-sample data [16]. In each case the in-sample period for model fitting and selection was specified from January 2007 to April 2011 (first 52 observations) while the out-of-sample period for forecast evaluation was specified from May 2011 to April 2012 (last 12 observations). All model comparisons were based on the results for the out-of-sample.