The above tests are based on a dispersion statistic U and a third moment statistic T. Small sample power comparisons are given for U and T. It is shown that Anscombe's tests are smooth in the sense of Rayner and Best (Ref. 3)· Comparisons are made with an empirical probability generating function test (Ref. 2). It is suggested not to use U, and that decisions be made based on bootstrap p-values rather than comparison with SE. It is shown that tests based on a fourth moment component of a smooth test statistic have good power. (12 refs.)
展开▼
