Long memory forecasting: An application to kse share index

Testing and estimating the long memory behavior in a given time series is one of the most crucial step to formulate the model. the simple rescaled range analysis r/s,and modified rescaled range analysis (mrs) are widely used techniques for this purpose. this paper attempts using mrs method for recognition and estimating the long memory to karachi stock index market. we employ autoregressive fractionally integrated moving average (arfima) model to capture the features of long memory in series. furthermore,a comparison is made between arfima and the one that take integer differencing i.e. autoregressive integrated moving average (arima). the two models are compared on the basis of some accuracy measures such as root mean square error (rmse) and root mean square percentage error (rmspe). the forecasting results show that both the estimated models provide approximately the same forecast values. it indicates that first-order arma process can be used as an effective alternative in forecasting kse share index. © 2014 pakistan journal of statistics.