The utilization of the Wilcoxon test to compare forecasting methods: A note

The selection of forecasting methods is a complicated issue. The situation is created by the need to identify the best method for a time series. Criteria for selection must be defined. If the sole criterion is accuracy, then a comparison of two forecasting methods can be made by testing them ex post while utilizing the time series. Care should be exercised in the selection of the accuracy metric. A statistical test should be made to determine if the results are significantly different. The test recommended is the Wilcoxson test since it is less restrictive to the nature of the error and less susceptible to outliers.