Gamma-minimax estimators in the exponential family
The Γ-minimax estimator under squared error loss for the unknown parameter of a one-parameter exponential family with an unbiased sufficient statistic having a variance which is quadratic in the parameter is explicitly determined for a class Γ of priors consisting of all distributions whose first two moments are within some given bounds. This generalizes the choice of Γ in Jackson et al. (1970) as well as the unrestricted case. It is shown that the underlying statistical game is always strictly...