A REMODELLING OF MONTHLY RAINFALL FREQUENCY IN UMUAHIA, NIGERIA, BY A BOX-JENKINS APPROACH
ETTE HARRISON ETUK *
Department of Mathematics / Computer Science, Rivers State University of Science and Technology, Nigeria.
IMO UDO MOFFAT
Department of Mathematics / Statistics, Uyo, Nigeria.
RICHARD CHINEDU IGBUDU
Department of Computer Science, Rivers State Polytechnic Bori, Nigeria.
*Author to whom correspondence should be addressed.
Abstract
Monthly rainfall frequency in Umuahia, Imo State, Nigeria, has recently been modelled as a SARIMA(0, 0, 0)x(0, 1, 1)12. This model was demonstrated to be better than the competing SARIMA(0, 0, 0)x(1, 1, 1)12 on all counts; that is, in terms of minimum AIC and maximum R2. Moreover in the latter model, the seasonal autoregressive, SAR, coefficient was observed to be statistically non-significant. This automatically made the former model suggestive. The statistical package used was R. However in this write-up by the use of Eviews 7 it is shown that the competing models are the same but the latter model is the better model on all counts including residual normal distribution. A wider 12-member array of likely SARIMA models is explored and the SARIMA(0, 0, 0)x(1, 1, 1)12 model is still found to be the best among them. Therefore forecasting of the series may be done using the SARIMA(0, 0, 0)x(1, 1, 1)12 model.
Keywords: Rainfall, frequency, time series, stationarity, SARIMA, R, Eviews 7