| Title |
Outlier search in arma models / |
| Translation of Title |
Nukrypusių reikšmių paieška ARMA modeliuose. |
| Authors |
Djaadi, Abdellah |
| Full Text |
|
| Pages |
33 |
| Keywords [eng] |
ARMA ; additive outlier ; innovative outlier ; BP ; Chang. |
| Abstract [eng] |
The purpose of this research is to study the problem of outliers in ARMA models and compare two outlier search techniques: Chang and BP methods. The aim of first method is based on an iterative procedure for parameters estimations then locating the additive and innovative outlier then distinguishing their types. The objective of second method is to generalize the BP outlier identification method for time series (ARMA). Our analysis shows that in most cases, the iterative procedure proposed by Chang is better than BP in detecting outliers. The latter method can not identify the type of declared outliers. However, there is some situation with adjacent outliers BP method slightly outperforms the existing techniques. |
| Dissertation Institution |
Vilniaus universitetas. |
| Type |
Master thesis |
| Language |
English |
| Publication date |
2022 |
