Unbiasedness as the Dual of Being Bayes
本文针对任意损失函数定义了决策函数的无偏性，这是Lehmann（1951）定义的推广，并证明该无偏性与贝叶斯性互为对偶关系，即交换随机变量与参数的角色后两者等价，并讨论了这一事实的若干推论。
In this note we define the notion of unbiasedness for a decision function for an arbitrary loss function. This is a generalization of Lehmann's (1951) definition. We show that this notion of unbiasedness is a dual to the notion of being Bayes; that is, if the role of the random variable and the parameter is interchanged, then unbiasedness is equivalent to being Bayes and vice versa. Some consequences of this fact are discussed.