The effects of additive outliers on time series components and robust estimation: A case study on the Oymapinar Dam, Turkey
Yükleniyor...
Tarih
2012
Yazarlar
Dergi Başlığı
Dergi ISSN
Cilt Başlığı
Yayıncı
Elsevier
Erişim Hakkı
info:eu-repo/semantics/openAccess
Özet
In this paper, the behaviour of the body of the Oymapinar Dam (Antalya,Turkey) resulting from reservoir water level changes was investigated by timeseries analysis. First, the possible best fitting methods for determining theexisting additive outliers in series were investigated. For this purpose, a timeseries with known parameters consisting of trend, periodical and stochasticcomponents was obtained. Then, a new time series with additive outlier(s) wascreated by adding outlier(s) to series with known parameters. The parameters ofthe components were determined by the least squares method and it was seenthat outlier(s) affect the components of the series. It was found that the correctdata were contaminated by an outlier to a similar degree as the autoregressivemodel. The outliers affecting the components were determined by Andrews,Bisquare, Cauchy, Huber and Welsch estimators in trend analysis. The resultsshow that the parameters of the trend were not affected by outliers and theresults produced by the estimators are numerically close to each other. Thus,series of the outlier(s) of the Oymapinar Dam were determined by Bisquareestimator in a trend analysis. As a result of series analysis it was determinedthat the linear trend resulted from a long-term periodical of changes and thedam body responded periodically depending on the water level changes
Açıklama
Anahtar Kelimeler
Time Series Analysis, Additive Outlier, Least Squares, M Estimation, Fast FourierTransform, Oymapinar Dam
Kaynak
Experimental Techniques
WoS Q Değeri
N/A
Scopus Q Değeri
Q2
Cilt
36
Sayı
3
