Article ID: | iaor20104966 |
Volume: | 29 |
Issue: | 4 |
Start Page Number: | 353 |
End Page Number: | 366 |
Publication Date: | Jul 2010 |
Journal: | Journal of Forecasting |
Authors: | Wang Zijun |
Keywords: | time series & forecasting methods |
Previous studies show that it is not always optimal to combine forecasts of alternative models. In this paper, we propose to use the recent advances in modeling directed acyclic graphs to study the issue of forecast combinations. In forecasting US unemployment rates, we demonstrate that the proposed procedure can be a useful tool for comparing information in rival forecasts and guiding the combination of individual forecasts.