Short Communication
Year: 2023 | Month: July | Volume: 10 | Issue: 7 | Pages: 950-958
DOI: https://doi.org/10.52403/ijrr.202307110
Improving the Reproducibility in Forecasting Research
P. Udhaya
Guest Lecturer in Mathematics, Jawaharlal Nehru Rajkeeya Mahavidyalaya, Port Blair, Andaman & Nicobar Island
ABSTRACT
The value of reproduction is recognized in many scientific fields. Reproducibility is a necessary condition for reliability because the inability to reproduce results indicates that methods are not sufficiently specified, thus preventing replication. This article describes how two independent teams of researchers attempted to reproduce the empirical findings of an important study, “Shrinkage estimators of time series seasonal factors and their effect on forecasting accuracy” (Miller & Williams, 2003, IJF). Teams of researchers proceeded systematically, reporting results before and after receiving clarifications. These inconsistencies have led to differences in conclusions about the conditions under which seasonal damping surpasses classical decay. The study addresses forecasting methods using a flow chart. It is argued that this approach to method documentation complements the provision of computer code by being accessible to a wider audience of forecasting practitioners and researchers. The importance of this research lies not only in its lessons for seasonal forecasting, but also in its approach to the reproduction of forecasting research.
Keywords: Forecasting practice, Reproduction, Seasonal Forecasting, Empirical analysis
[PDF Full Text]