ABSTRACT
Time series that have seasonal effects with long periods relative to the observation frequency can exhibit spurious long memory. The effect of these seasonalities on the periodogram is similar to that of structural breaks and non-periodic trends, but it only influences the seasonal frequencies and their harmonics. Still, the effect causes a sizable bias of popular estimators such as the local Whittle estimator. To overcome this, we propose a robust local Whittle estimator based on the omission of the affected periodogram ordinates. In a Monte Carlo study, we compare this estimator with a robust log-periodogram regression-based estimator known in the literature. An application to electricity load series demonstrates the potential of robust estimators for empirical research.
Acknowledgments
We would like to thank the referee for detailed comments that helped us to improve the manuscript substantially. Financial support of the Deutsche Forschungsgemeinschaft (DFG) is gratefully acknowledged.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1 Dataset from https://data.open-power-system-data.org/time_series/.