On tests of independence in a trivariate exponential distribution (Q792716): Difference between revisions

The author extends the \textit{G. K. Bhattacharyya} and \textit{R. A. Johnson} procedure [J. Am. Stat. Assoc. 68, 704-706 (1973; Zbl 0271.62029)] (developed for the bivariate case) to the trivariate exponential distribution. This is achieved by transforming the independence hypothesis to one concerning the equality of scale parameters of three exponential distributions. The tests are conditional tests. The two procedures suggested are the Sukhatme LR test and the Hogg and Tanis iterated test. In the first case the asymptotic distribution is shown to be chi square. In the latter case the author gives the expression for the exact power of the test. The minimal sufficient statistics and their distributional properties are also discussed.