Hata terimlerindeki otokorelasyonun Dickey Fuller birim kök testinin gücü üzerindeki etkisi: Simülasyon model yaklaşımı
Files
Date
2017-09-07
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Uludağ Üniversitesi
Abstract
Bu çalışmada en temel birim kök testi olarak kabul edilen Dickey Fuller birim kök testlerinin, hata terimlerinin korelasyonlu olması sonucu etkilenmesi durumu incelenmiştir. Bilindiği üzere Dickey Fuller birim kök testleri ile sistemdeki otoregresif değişkenin parametre katsayısının 𝜏 (tau) istatistiği ile söz konusu değişkenin durağanlığı belirlenmektedir. Ancak parametre katsayısının bire yakın çıkması durumunda test eleştirilmektedir. Test bu durumda durağanlık olgusu yerine durağan dışılığı sıkça önermektedir. Çalışmada belirlenen sonuçlara göre, hata terimlerinin korelasyonlu olması ve bu durumunun göz ardı edilmesi ile durağan seriler daha sık bir şekilde durağan dışı olma eğilimindedir. Bununla beraber otokorelasyonun zayıf olsa da sistemden uzaklaştırılması halinde serinin durağan çıkma olasılığı daha net bir şekilde önerilmektedir.
In this study, as the most basic unit root test Dickey Fuller tests, which effects are considered, have been investigated with including error term autocorrelation. As known, the Dickey Fuller Unit Root tests determine the stability of the variable by using the 𝜏 (tau) statistic of the autoregressive variable coefficient parameter in the system. However, if the parameter coefficient is close to one, the test is criticized justly. In this case test suggests frequently non-stationary rather than stationary. The aim of the study is to discuss the determination of the stability under the correlation of error terms. At the study the stationary test is discussed with the simulation results if there is a case the error terms are correlated. According to the results determined in the study, the stationary series were being suggested more often as nonstationary, considering the correlation of error terms. However, even if autocorrelation is weak, when it is removed from the system, being stationary has become more clearly suggested if series truly are. As a result, even if there is autocorrelation that can be accepted as insignificant in the system, it is decided that to be removing from the system in order to judge being stationary accurately.
In this study, as the most basic unit root test Dickey Fuller tests, which effects are considered, have been investigated with including error term autocorrelation. As known, the Dickey Fuller Unit Root tests determine the stability of the variable by using the 𝜏 (tau) statistic of the autoregressive variable coefficient parameter in the system. However, if the parameter coefficient is close to one, the test is criticized justly. In this case test suggests frequently non-stationary rather than stationary. The aim of the study is to discuss the determination of the stability under the correlation of error terms. At the study the stationary test is discussed with the simulation results if there is a case the error terms are correlated. According to the results determined in the study, the stationary series were being suggested more often as nonstationary, considering the correlation of error terms. However, even if autocorrelation is weak, when it is removed from the system, being stationary has become more clearly suggested if series truly are. As a result, even if there is autocorrelation that can be accepted as insignificant in the system, it is decided that to be removing from the system in order to judge being stationary accurately.
Description
Keywords
Zaman serisi modelleri, Birim kök, Otokorelasyon, Parametrik modeller, Simülasyon, Time series models, Unit root, Autocorrelation, Parametric models, Simulation
Citation
Çınar, M. ve Hepkorucu, A. (2015). "Hata terimlerindeki otokorelasyonun Dickey Fuller birim kök testinin gücü üzerindeki etkisi: Simülasyon model yaklaşımı". International Journal of Social Inquiry, 8(1), 29-61.