Bayes estimates of estimable parameters against Dirichlet prior processes and squared error loss are obtained in one and two sample cases. Under certain conditions on kernels the limits of Bayes estimates have the asymptotic properties similar to the U-statistics. Furthermore some asymptotic properties are examined.
This paper extends the results of Berger [3] by Efron and Morris [5] approach and specifies a class of minimax estimators for β in a linear regression model y=Xβ+u with loss (_??_-β)'Q(_??_-β)/σ2 where Q is arbitrarily fixed. A sufficient condition for a minimax estimator in the class to be minimax under a different quadratic loss is also given.
Attempts are made to express the estimators and their residuals of sub- and redge-regressions in terms of the eigen values and vectors of the correlation matrix of the dependent and independent variables.
This paper deals with the classification statistic W* in covariate discriminant analysis with two multivariate normal populations. An asymptotic expansion of the distribution of the Studentized classification statistic W* is given with an error of the third order of the reciprocal of the number of observations. The expansion will be useful in setting our cut-off point to achieve a specified probability of misclassification.
This paper provides practical bounds for the number of treatments common to any two blocks of singular group divisible, triangular, and L2 symmetrical partially balanced incomplete block designs. Some examples attaining these bounds are presented.