Fixed-Size Confidence Regions for the Multinormal Mean in an Intraclass Correlation Model

SYNOPTIC ABSTRACT

The problem of constructing fixed-size confidence regions for estimating the mean vector of a multivariate normal distribution is considered when the covariance matrix has the intraclass correlation coefficient. Two-stage procedure and purely sequential procedure are proposed. In order to apply these procedures, an asymptotic approximation to a distribution is used.

Key Words and Phrases:

• multivariate normal distribution
• intraclass correlation model
• confidence region
• asymptotic expansion
• two-stage procedure
• sequential procedure