BISPECTRA AND LINEARITY OF HYDROLOGIC TIMES SERIES

The common assumption in modeling hydrologic time series is that the series is linear. However, many real time series cannot be adequately described by linear models. In recent years, several researchers have developed tests based on bispectral density function which can be used for testing the linearity of time series. Two tests of linearity of time series proposed by Rao and Gabr (1980) and Hinich (1982) based on the bispectrum of time series are investigated and used in this study. The following aspects of bispectral analysis are also investigated: the accuracy of estimation, the effects of sample size variation on estimation, and the assumption of normal distribution of the estimates. The results indicate that these tests can detect the presence of non-Gaussianity and nonlinearity in time series. Also, Gasser test give the adequate preliminary information about the symmetry of time series