Authors: KADRİ YÜREKLİ, AHMET KURUNÇ, FAZLI ÖZTÜRK
Abstract: In ARIMA modeling studies, the selection of a best model fit to historical data is directly related to whether residual analysis is performed well. Therefore, diagnostic checks including the independence, normality and homoscedasticity of residuals is the most important stage of ARIMA model building. This study is concerned with testing residuals from ARIMA models for monthly streamflow data from the Çekerek Stream watershed. Alternative tests including the Ljung-Box Q statistic, runs test and turning point test for independence analysis of the residuals; Kolmogorov-Smirnov and Anderson-Darling tests for normality of residuals; and Goldfeld-Quandt, Breusch and Pagan and Spearman's rho approaches for the homoscedasticity of residuals were used. The selected parsimony model for each data set among the ARIMA models fulfilled the diagnostic checks, considering the Schwarz Bayesian criterion.
Keywords: ARIMA model, Monthly streamflow, Çekerek stream, Diagnostic checks
Full Text: PDF