The Use of Maximum Entropy to Estimate Input‐Output Coefficients From Regional Farm Accounting Data
提出用广义最大熵方法估计农场投入成本在不同产出间的分配系数，并与最小二乘、贝叶斯和线性规划比较，发现该方法能处理奇异性和零观测等问题，对农业经济研究者有实用价值。
This paper proposes the use of the Generalised Maximum Entropy (GME) method to estimate input‐output coefficients, which reflect the unobserved allocation of farm input accounting costs to the various outputs produced. The GME method uses Shannon's information criterion as a basis for estimation. The performance of the GME method is compared with three other estimation techniques: Ordinary Least Squares (OLS), Bayesian estimation, and Linear Programming (LP). The various methods are applied to accounting data from a sample of beef‐dairy farms in Brittany, France. The analysis shows that the GME method offers an interesting alternative to “traditional” estimation methods. In contrast with the latter, though, the GME method is suitable to handle easily the problems of singularity, constrained estimation, and zero‐observations. Moreover, due to its flexibility, transparency and relative ease of implementation, the GME method is of great value to practitioners. However, the sensitivity of the GME estimates with respect to the design of the prior information set needs to be investigated further.