Zaman serilerinde sapan değer tarama süreçlerini modelleme ve performans analizi
Date
2003
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Uludağ Üniversitesi
Abstract
Sapan değerler, Parametre tahminlerinin yanlı olmasına neden olan, kişilerin, aletlerin hatalarından veya hatalı kullanımları ile oluşabilen ya da doğal rasgelelik sonucunda ortaya çıkabilen az sayıda gözlem değeridir. Geçmişte tanıma dayalı olarak tespit edilebilen sapan değerler, günümüzde özellikle ARIMA modellerde otokorelasyona dayalı yöntemler ile tespit edilmektedir. Bu çalışma zaman serilerinde, aynı anda test yöntemleri olarak bilinen LS (Least Square), M-H (Method of Huber-Type), M-B (Method of Bisquare-Type), GM-H (Method of Generalized Huber Type), GM-B (Method of Generalized Bisquare-Type) ve ITERATIVE (Ardışık) yöntemlerin performansları, sapan değer sayıları (1, 2 veya 3), ve sapan değer türleri (AO ve IO) özel seçimli, çok-etkenli (2x3x6) deney düzeni için tasarlanmış ve varyans analizi tekniği ile test edilmiştir. Araştırma sonucunda etkenler önemli bulunmuştur.
Investigated in this study which is called outliers have been regarded as the observations which are extremely different from the others in a sample one, two or more than two observations which effect the estimation and forecasting process in a great extent. The data used in this study were obtained from the simulation experiments performed on several time series model by Chang, Tiao and Chen The original simulation results were reorganized and remodeled under a suitable experimental design. Two different time series outliers types (AO, IO) in three different size (1, 2 or 3) and six different detection methods (LS, M-H, M-B, GM-H, GM-B and iterative) were arranged. The type of factorial design was that is 3 factors each has 1 levels and 1 factor at 3 levels each. The result of analysis of variance performed for this design indicated that the main effect except that of sensitivity coefficient were statistically significant.
Investigated in this study which is called outliers have been regarded as the observations which are extremely different from the others in a sample one, two or more than two observations which effect the estimation and forecasting process in a great extent. The data used in this study were obtained from the simulation experiments performed on several time series model by Chang, Tiao and Chen The original simulation results were reorganized and remodeled under a suitable experimental design. Two different time series outliers types (AO, IO) in three different size (1, 2 or 3) and six different detection methods (LS, M-H, M-B, GM-H, GM-B and iterative) were arranged. The type of factorial design was that is 3 factors each has 1 levels and 1 factor at 3 levels each. The result of analysis of variance performed for this design indicated that the main effect except that of sensitivity coefficient were statistically significant.
Description
Keywords
ARIMA, AO, IO, Outlier, Detection methods, Zaman serileri
Citation
Kaya, A. (2003). "Zaman serilerinde sapan değer tarama süreçlerini modelleme ve performans analizi". Uludağ Üniversitesi İktisadi ve İdari Bilimler Fakültesi Dergisi, 22(1), 271-279.