National Technical University of Athens (NTUA)
Ανάπτυξη στατιστικών μεθόδων σε μοντέλα χρονοσειρών για την ανίχνευση απομονωμένου σημείου ή σειράς
Abstract
dc:descriptionOutlier analysis in time series has an interesting interpretation. First of all, outliers can help in fitting imperfect theories to complex real world phenomena. On the other hand, outliers can also be used to reveal where a theory does not work, or to check what aspects of it need refining in order to better describe the real world. The examination of outliers can therefore be justified not only from the traditional data analysis perspective, but also by appealing to the interaction of theoretical and empirical point of view. Through this thesis some new methods of detecting an outlier in some complex time series models are examined. Concerning a set of time series, an interesting aspect is whether any series is an outlier in relation to the others. So, some models for sets of time series are described and tests are derived in order to detect if the mean level of a series is an outlier. The performance of these tests is compared to simple approximate tests based on standard outlier tests. Furthermore, the problem of detecting outlying observations in a set of time series is examined. Finally, we discuss the presence of an outlier in models for time series of counts. The methods of detection are necessary because the presence of outliers may affect the final results of statistical analysis (for example, outliers may bias our estimates or cause problems in forecasting).
Degree
thesis:*- Grantor dc:publisher
- National Technical University of Athens (NTUA)
- Year dc:date
- 2002
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Karioti, Vassiliki
- Καρυώτη, Βασιλική
Subjects
dc:subject × 16- Απομονωμένα σημεία
- Χρονοσειρές
- Απομονωμένες σειρές
- Μοντέλα παλινδρόμησης
- Χρονοσειρές απαρίθμητων δεδομένων
- Έλεγχος λόγου πιθανοφανειών
- Outliers
- Time series
- Outlying series
- Observation-driven regression
- Models for time series of counts
- Likelihood ratio tests
- Φυσικές Επιστήμες
- Μαθηματικά
- Natural Sciences
- Mathematics
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/16647