1: \begin{abstract}

2: Covariance matrix plays a central role in multivariate statistical analysis.

3: Significant advances have been made recently on developing both theory and

4: methodology for estimating large covariance matrices. However, a minimax

5: theory has yet been developed.

6: %Existing estimation methods and theoretical analyses employ the

7: %strategy of reducing the matrix estimation problem to that of

8: %estimating vectors, with the aim of estimating each row/column

9: %optimally.

10: In this paper we establish the optimal rates of convergence for estimating

11: the covariance matrix under both the operator norm and Frobenius norm.

12: It is

13: shown that optimal procedures under the two norms are different and

14: consequently matrix estimation under the operator norm is fundamentally

15: different from vector estimation. The minimax upper bound is obtained by

16: constructing a special class of tapering estimators and by studying their

17: risk properties. A key step in obtaining the optimal rate of

18: convergence is

19: the derivation of the minimax lower bound. The technical analysis requires

20: new ideas that are quite different from those used in the more conventional

21: function/sequence estimation problems.

22: \end{abstract}