|
Abstract:
Four purposes are set for this paper. The first one is for the seismic pattern identification of monthly number of earthquakes of three counties on the eastern coast of Taiwan - Yilan, Hualien and Taitung. The monthly number of earthquakes were obtained from the seismic archive of the Central Weather Bureau (CWB) of Taiwan from January 1995 to June 2017, totally 22.5 years (270 months).The versatile autoregressive integration moving average (ARIMA) model is used to identify seismic patterns in these three areas. ARIMA(3,0,0), ARIMA(2,0,1), and ARIMA(1,0,1) models can completely satisfy the statistical theories testing procedures for Yilan, Hualien and Taitung respectively. The second purpose of this paper is to find out whether the time series of monthly number of earthquakes in the past 270 months in Yilan, Hualien and Taitung are stationary or not. The Augmented Dickey-Fuller (ADF) unit-root testing method is used to check the stationarity of the time series, and the results show that Yilan, Hualien and Taitung all reveal to be stationary. The third purpose is to check whether the time series of the monthly number of earthquakes in Yilan and Taitung are cointegrated with Hualien or not. The outcome shows the time series of Hualien are cointegrated with Yilan and Taitung, and no spurious effect exists in regression analyses. The regression equations show that a unit increase of earthquake in Hualien may result in an increase of 0.2151 unit earthquake in Yilan, and 0.1692 unit increase of earthquake in Taitung. The fourth purpose is to find out whether the autoregressive conditional heteroskedastic (ARCH) variances of the monthly number of earthquakes in the three counties have ARCH(1) property or not. The ARCH test shows the volatile variances of time series for both Hualien and Yilan are heteroskedastic, but Taitung is homoskedastic.
|
