Pages 24

Prior to the interest in tests specific to identifying non-normality, initiated at least in part by the test of Shapiro and Wilk (1965), distributional assumptions were verified most often by more general goodness of fit tests which could be used for any simple null hypothesis FQ(X). These tests are unique from the tests for normality described in the preceding chapters in that, under a simple hypothesis, their distribution is independent of the null distribution and therefore tables of critical values are identical regardless of the hypothesized null distribution. However, prior knowledge of distributional parameters is relatively rare in actual practice.