Search

All the studies regarding time series methods are useful only in case the series in interest do not display seasonal patterns. That is why it is of great importance to take the time series properties of the series like seasonal patterns or trends into account while dealing with economic time series data and the research on what form of seasonality exists (deterministic or stochastic) in the data in interest and thus the way of modelling seasonality is also crucial (Türe & Akdi, 2005, p.3). Considering its importance with this respect, in this application, it has been aimed to decide about which seasonal pattern quarterly GDP (Gross Domestic Product) series displays over 1998Q1-2014Q4 for Turkey by recoursing to DHF (Dickey, Hasza and Fuller) and HEGY (Hylleberg, Engle, Granger and Yoo) test procedures and it has been mainly focused on the dummy variable and trigonometric representations of deterministic seasonality.

Bu çalışmada Türkiye'nin 1995.'01--2015:03 dönemine ilişkin aylık enflasyon oranları için en iyi Mevsimsel Otoregresif Bütünleşik Hareketli Ortalama (SARIM4)' modelini bulmak amaçlanmıştır. Box Jenkins metodolojisine dayalı model tanımlamasından önce HEGYmevsimsel birim kök testi uygulanmıştır. Mevsimselfark alma dereceleri OCSB ve CH testleri kullanılarak saptanmıştır. Son olarak, sırasıyla adımsal (stepwise) ve adımsal olmayan (non-stepwise) seçim yöntemleri kullanılarak seçilen sürüklenmeli ARIM4(1,1,1)(1,0,2)[12] ve ARIIMA( ,1 ,I )(2,0, 0)[ modelleri karşılaştırılmıştır. Sonuçlar adımsal yöntem kullanılarak seçilen sürüklenmeli ARIM4(1,1,1)(1,0,2)[12] modelinin en iyi uyan SARIIWA modelini belirlemede adımsal olmayan modele göre daha iyi olduğunu göstermiştir. In this study, it has been aimed to find the best ‘Seasonal Autoregressive Integrated Moving Average (SARIMA) modelfor monthly inflation rates for Turkish economy over the period 995:01 -201 :03 Before the model identification based on Box Jenkins methodology, HEGY monthly seasonal unit root test has been applied. The orders of seasonal difi‘erencing have been detected through OCSB and CH tests. Finally, ARIM4(1,1,1)(1,0,2)[12] with drift model chosen by using stepwise selection method and ARIMA(1,1,1)(2,0,0)[12] with drift model chosen by using non-stepwise selection have been compared. The results have shown that the former model is better as the bestfitted SARIMA model.