| Bibtype |
Inproceedings |
| Bibkey |
Fried/Raabe/2012a |
| Author |
Fried, Roland and Raabe, Nils and Thieler, Anita |
| Editor |
Colubi, Ana and Fokianos, Konstantinos and Kontoghiorghes, Erricos John and González-Rodríguez, Gil |
| Title |
On the robust analysis of periodic nonstationary time series |
| Booktitle |
Proceedings of COMPSTAT 2012 |
| Series |
Proceedings of the COMPSTAT |
| Pages |
245--257 |
| Publisher |
The International Statistical Institute/International Association for Statistical Computing |
| Abstract |
Robust methods for the detection of periodic signals in time series are presentedand discussed. The motivation are applications to the modeling of drilling processes and in astroparticlephysics. The considered signals which are measured during drilling processes consistof equidistant observations with known period and gradually changing periodic structure. Thebasic objective is to understand the granularity of di erent materials by analyzing the periodicstructure in order to design suitable simulation models, which render subsequent optimizationof the system possible. For this aim, robust nonparametric smoothers and edge detectors for thereconstruction of periodic jump surfaces are developed and combined. In astroparticle physics,the situation is worse because of very irregular observation times and heteroscedastic noise. Themain interest is in the detection and identication of periods, if they exist at all. Suitablymodied robust nonparametric smoothers allow construction of generalized periodograms. Signicant periods are identied by application of rules for outlier detection to such periodograms.All methods are investigated by applications to simulated and real data with and without outliers.
|
| Year |
2012 |
| Projekt |
SFB876-C3 |
