NONLINEAR TIME SERIES MODELLING OF DAILY MEAN TEMPERATURE IN NIGERIA

Abstract

ABSTRACT This- study is based on nonlinear time series modelling in Nigeria. Two models were considered. The first model adopted Gallant parametric model with trigonometric transformation augmented with polynomial of order two, to address the problem of missing values. This is called Augmented Nonlinear Parametric Time Series Model (ANPTSM). In addition, the second model was formulated using Ordinary Least Squares Method, trigonometric transformation and method of calculating an expected value in contingency table of multivariate observations. This is called Modifie~ Nonlinear Trigonometric Transformation Time Series Model (MNTTSM) . • The ANPTSM IS a nonlinear parametric time senes model usmg trigonometric transformation to solve the problem of missing values in time series, while MNTTSM is a model for the conditional mean of modified nonlinear time series. The consistency and efficiency of the models formulated were determined. The study examined Nonlinear Trigonometric Transformation as well as Augmented Nonlinear Trigonometric Transformation with polynomial of order two. The two models were practically tested and compared using daily mean of temperature for Ikeja, Ibadan, Ilorin, Minna and Zaria with different rates of missing values. The results were used to determine the consistency and efficiency of the models formulated which showed that MNTTTSM is able to address the problem of missing values and forecast better than the ANPTSM.