Prediction in production economics helps individuals and groups to make choices among the uses of resources. Prediction is usually mad by using a sample drawn from a population in which the variables quantities, or parameters to be estimated are already in existence. But in many situations with which biological and economic research workers deal, the parameters they want to predict are not found in an existing population because farmers have not yet used the recommended production techniques. They therefore frequently must formulate predictions either by inference from a sample that is assumed to represent a population, or by budgeting procedures. Conventional budgeting procedures frequently have limited usefulness because the empirical data are assumed to be discrete, linear, and without error. This article brings out some refinements in conventional budgeting procedure. These suggested modifications are believed to increase the information that can be gained from a smaller number of budgets, and to give greater knowledge of the predicational error involved.