BAYESIAN FORECASTING USING NONLINEAR TIME SERIES MODELS

Abstract

It is generally considered that the statistical forecasting methods are superior to the methods which are based on non statistical principles. Non statistical methods are less sophisticated and simple to understand for an ordinary person. This paper presents the forecast based on purely statistical methods called Bayesian forecasting. Asymmetric time series method have been used along with Kalman s Filter results. GARCH process has been used to estimate the nonconstant variances in the observation equation and in the system equation. Finally the exchange rate of Pakistani rupee against U.K. pound is used for the forecasting purpose.