非平衡单向分类模型中方差分量的Stein型区间估计

The interval estimation for variance components using Stein method in unbalanced one-way random model

在非平衡单向分类模型中,常规方差分量的区间估计都是以组间平方和与组内平方和为统计量,通过引入包含组间方差分量信息的统计量USS,将均值与平方和的线性组合作为新的统计量.改进后的区间估计与传统的最小长度区间估计相比,长度不变但置信度更高.

The problem of improving interval estimators of variance components was considered in the unbalanced one-way randomly classified model.By introducing the statistic USS containing the information of the variance component between groups,the linear combination of the mean and the sum of squares was used as a new statistic.It has been shown that interval estimates of variance components rely on not only sums of squares but also the population mean in the basis of the above construction.Compared with the traditional minimum length interval estimation,the improved interval estimation has the same length but bigger confidence.