Abstract
We consider an asymptotic second-order efficiency of a two-stage procedure which was proposed to construct a fixed-width confidence interval for a linear function of normal means. It is shown that no matter how the initial sample size is chosen, the two-stage procedure does not become asymptotically second-order efficient. However, if it is possible to assume a known lower bound for each unknown variance, it is shown that a proper choice of the initial sample size enables us to make the two-stage procedure asymptotically second-order efficient.
Acknowledgment
The author wishes to thank the referees for their helpful comments.
Notes
Recommended by P. Chen