Abstract
This paper focuses on confidence estimation for constrained parameters. It is shown how the method of tail functions can be used to engineer a constrained confidence interval which is optimal in terms of prior expected length. The strategy is compared with the unified approach of Feldman and Cousins (1998) and illustrated by application to inference on the normal mean.
Original language | English |
---|---|
Pages (from-to) | 134-140 |
Number of pages | 7 |
Journal | Mathematical Scientist |
Volume | 33 |
Issue number | 2 |
Publication status | Published - 2008 |