Abstract
This study proposes two methods for detecting outliers in functional time series. Both methods take dependence in the data into account and are based on robust functional principal component analysis. One method seeks outliers in the series of projections on the first principal component. The other obtains uncontaminated forecasts for each data set and determines that those observations whose residuals have an unusually high norm are considered outliers. A simulation study shows the performance of these proposed procedures and the need to take dependence in the time series into account. Finally, the usefulness of our methodology is illustrated in two real datasets from the electricity market: daily curves of electricity demand and price in mainland Spain, for the year 2012.
Keywords
- Functional data analysis
- functional principal component analysis
- functional time series
- outlier detection
- electricity demand and price
Rights
Copyright
From February 2013 articles are under a Creative Commons license: CC BY-NC-ND You must attribute the work in the manner specified by the author or licensor (but not in any way that suggests that they endorse you or your use of the work), you may not use the work for commercial purposes and you may not alter, transform, or build upon the work.
