Forecasting With The Theta Method

4Empirical Applications with the ‐method

By Dimitrios D. Thomakos and Kostas I. Nikolopoulos

4.1 Setting up the Analysis

4.1.1 Sample Use, Evaluation Metrics, and Models/Methods Used

In this chapter, we go through several real‐life series and present empirical applications for the material of the previous chapters. The aim here is to examine not only the relative forecasting performance of the proposed methods, against some well‐known benchmarks, but also to understand how the forecasting performance of the ‐ or ‐based forecasts changes when we change the data‐generating process (DGP), the trend function, and other parameters used in the forecast functions of the methods. Our approach is the standard rolling window training/evaluation approach that appears in the literature, where we generate genuine, out‐of‐sample rolling forecasts and then evaluate them over a number of periods.

Consider therefore a sample split of total observations. We say that is the rolling window on which we estimate ...