Improved Confidence Intervals for the Variance of a Normal Distribution

In considering the problem of confidence estimation for the variance of a normal distribution, Tate and Klett [5] raise two questions. “Does the interval of shortest length based on (, S2) depend only on ” (S2 denotes the sum of squared deviations.) “Among those intervals based only on S2 is the interval of shortest length necessarily of the form ” The answer to the first question is no; to the second, yes. The results are given for the general linear hypothesis model of rank p. The amount of improvement of new intervals is discussed.