Опубликовано на портале: 24-06-2004Charles S. Davis, Michael A. Stephens Applied Statistics. 1989. Vol. 38. No. 3. P. 535-582.
Empirical distribution function (EDF) statistics for goodness of fit are based on a comparison of the hypothesized distrribution function F(x) with the empirical distribution function Fn(x). When F(x) is continuous and completely specified, it has long been known that? in general, EDF statistics give more powerful tests of Ho then the classical X2 test. This work has made possible to use EDF statistics very easily when F(x) is completely specified and also for two practical situations.