Interval estimation of noncentrality parameter of F-distribution

Powers of statistical tests of hypotheses most often require noncentral chi(2), t or F-distributions. The noncentrality parameter (NCP) is a function of the sample size. In the process of designing experiments or, more specifically, planning a clinical trial, stating the proper sample size that would with specified power enable the detection of a desired difference, is an essential task. Ideally, a pilot study or other source of information on the value of test statistic in a similar setting is available. Hence estimation of the NCP becomes an important issue. Here we present a simple approach to construction of a confidence interval of an NCP of an F-distribution and compare the derived interval with a different type suggested in the literature.