2

I am confused about this. Are they the same? There are some problems in which I am asked to find the Uniformly Minimum Variance Unbiased Estimator (UMVUE) first and then check if it achieves the CRLB. Yet from what I've read, I understand that an unbiased estimator that achieves CRLB is UMVUE. Thanks.

user164144
  • 1,077
  • 7
  • 18

1 Answers1

4

They are not the same. Achieving the CR lower bound is a sufficient condition for an unbiased estimator to be UMVUE. However, it is not necessary.

For example, Example 3.10 in this link gives an estimator that is UMVUE (by the Lehmann-Scheffe theorem) but does not attain the CR Lower bound.

Flowsnake
  • 783
  • 3
  • 12
  • Is there a general procedure for finding UMVUEs without the use of CRLB? – user164144 Jul 06 '17 at 01:02
  • 1
    @user164144 Yes, if the unbiased estimator is a function of a complete sufficient statistic, then it is the UMVUE. This is the Lehmann-Scheffe theorem. In particular, let $T$ be a complete sufficient statistic and $S$ an unbiased estimator. Then $E[S | T]$ gives the UMVUE. – Flowsnake Jul 06 '17 at 01:03