A0415
Title: An improved test for uniform superior predictive ability
Authors: Verena Monschang - University of Muenster (Germany) [presenting]
Mark Trede - University of Muenster (Germany)
Wilfling Bernd - University of Muenster (Germany)
Abstract: The test for uniform superior predictive ability (uSPA) is analyzed, demonstrating that it does not always keep its nominal size. Simulations show that the testing procedure is not able to control the type I error rate. An improved testing approach that keeps the nominal size asymptotically is proposed. Monte Carlo simulations investigate the test's power properties. As an empirical illustration, we compare the predictive ability of two forecasting models.
