Abstract

                                In this paper two new methods are attempted for forecasting in agriculture through interval estimation. The first method relaxes the assomption of constant variance for residual errors, which is used in the usual least squares method in curve fitting. The second method involves fitting of three growth curves for normal. above normal and bolow normal values of the time series data. The above methods were illustrated along with the usual least-squares for the data of Production and Productivity of Paddy for Andhra region in Andhra Pradesh from 1961-62 to 1980-81. The first method gave lower and upper boundaries with minimum bias as compared to other two methods.

Key words : Least squares method, Growth curves, Interval estimation