Confidence Intervals for a Multivariate Normal Mean in the Case of Empirical Bayes Estimation Using Pearson Curves and Normal Approximations

摘要

More accurate confidence intervals for sample baseball data were obtained by using Pearson curves instead of the normal approximations. The theory and results for the Stein estimator for the vector case were also extended to that for the matrix case and confidence intervals were obtained. The unbiased risk estimator was also calculated.