Forcasting Acha production in Nigeria: A comparison of grafted polynomial and linear models

Abstract

Paucity of functional forms is one of the problems associated with forecasting economic time series data because not all economic time series correlate linearly to trend over an entire sample period. This paper attempt to investigate the performance of grafted polynomials in forecasting Acha production in Nigeria. The observed trend in Acha production in Nigeria from 1970-2004 provides the database for the study. A grafted model (Quadratic-Quadratic-Linear) was estimated to forecast Acha output for periods outside the sample. The estimated structural parameters gave an R squared values of 0.81 and 0.99 for the linear and grafted respectively. Forecast were generated to evaluate the predictive performance of the estimated models, the grafted model performed better than the linear because the forecast were closer in values to the observed values.Theil inequality coefficient of 0.015714 and 0.129091 was obtained for the grafted and linear respectively, this further reinforces the superiority of the grafted polynomial. It is better therefore to fuse the entire local trend in forecasting of this nature to achieve a better time series prediction.br>
Journal of Agriculture, Forestry and the Social Sciences Vol. 3 (2) 2005: pp.42-50