Methods of Estimating the Input Coefficients for Linear Programming Models
讨论如何从样本数据估计线性规划模型中的非负投入系数，比较了最小化绝对偏差均值、最小化最大绝对偏差和不等式约束最小二乘等方法的优劣，适用于农业产品组合优化问题。
Abstract In order to apply the linear programming method to determine the most profitable product mix of a farm, it is often necessary to estimate the input coefficients from sample data. The estimated coefficients must all be nonnegative. Estimates from regression models of input demand equations are not always nonnegative. The sample average of inputs used per acre of various crops provides nonnegative estimates but may suffer from selectivity bias. In this study the methods of minimizing the mean of absolute deviations and of minimizing the maximum absolute deviation are discussed as alternatives to inequality‐constrained least squares.