Abstract:
The problem of quadratic risk minimization is solved for the componentwise shrinkage estimation of expectation value vectors for normal vectors with independent components and unit variance. An unimprovable shrinkage function is found, and a shrinkage estimator of the expectation value vector is constructed unimprovable with accuracy to terms that are small for large dimension and large sample size.
Keywords:shrinkage estimator, dominating estimators, large dimension.
Fulltext:
PDF file (934 kB)