A1421
Title: Forecast evaluation with functional data
Authors: Jack Fosten - King's College London (United Kingdom)
Piotr Kokoszka - Colorado State University (USA)
Tim Kutta - Aarhus University (Denmark)
Shixuan Wang - University of Reading (United Kingdom) [presenting]
Abstract: The aim is to propose methods for comparing the accuracy of two competing sets of functional forecasts. This is increasingly relevant as many economic and financial forecasters are interested in predicting variables which are observed as functional data objects. However, to date, there have been no formal statistical tests to evaluate the relative accuracy of competing functional forecasts. A suite of novel tests is proposed, building on the classic Diebold-Mariano test, to provide formal statistical guidance on forecast accuracy in the case of functional data. The asymptotic properties of the tests are derived, as well as self-normalized versions, and the validity of analytic or bootstrap-based critical values is demonstrated. The finite sample performance of the tests is investigated using Monte Carlo simulations. The practical usefulness of the tests is demonstrated by evaluating competing forecasts of U.S. yield curves based on forward rates and a naive benchmark.
