Selecting a minimax estimator doing well at a point
Selecting a minimax estimator doing well at a point
复制标题
选择在某个点表现良好的极小极大估计器
DOI:
10.1016/0047-259x(86)90091-6
复制
发表时间:
1986
影响因子:
1.6
通讯作者:
Z. Zheng
中科院分区:
文献类型:
--
作者:
Z. Zheng
Let Xi-N (ei, I,), i= 1, 2,..., n (n> p+ 1, p> 1) be an independent finite sequence of p-dimensional vectors. Let X=(X,,..., X,,) be p xn matrix of random variables with the matrix of means 6=(or, e,,..., 0,). Efron and Morris [2] generalized the James-Stein estimator: they proved that X-(np-1) SC’X (along with its positive part (I-(np-1) Sl)+ X) is a minimax estimator. Stein [3] considered the estimators of form X+ V lnf (l), where I=(II, I,,..., Z,)’is a vector with components I,> Z2>.** lp 2 0, the characteristic roots of S= XX’, and V lnf (f)=((a/&Y,) lnf (Z)) is the matrix of partial derivatives of the function In f (Z). He gave a condition under which the corresponding estimator X+ V Inf (l) is minimax. Zheng [4] gave a class of minimax estimators of the form X+ V lnf (l), where