Evaluation of Alternative Flexibility Restraint Procedures for Recursive Programming Models Used for Prediction

Abstract

AbstractRecursive programming models for estimating production response may be generalized from the original “recursive” formulation to a formulation that simply attempts to limit the linear programming solution to a reasonable subset. This article outlines a framework for evaluating the effect of alternative flexibility restraint procedures on the accuracy of such recursive programming models. Statistical estimates are developed for the total expected error of predictions, given a specified set of population and model characteristics. The analysis suggests that recursive programming models achieve some measure of statistical accuracy.