我看完这几个例子之后,觉得没有自相矛盾。杨重法的203页的例子,以及Ann Lehman 302页的例子,都是samll sample size的例子。Ann Lehman在 讨论302页的例子时就指出了此时运用sphericity test 有倾向 falsely rejct the null hypothesis of significant departure from the homogeneity of variance assumption, 但是当你只勾选Test Each Column Separately Also时报告的multivariate statistics is good enough to complete the test. 希望以上能回答你的疑惑。