many stats packages provide the Shapiro-Wilk test for normality
Shapiro, S. S., & Wilk, M. B. (1965). An analysis of variance test for
normality. Biometrika, 52, 591-611.
other commonly provided statistics are the kolmogorov-smirnov or its
lillefors variant. skew and kurtosis are also often given (although tests
based on these statistics would be (a) biassed and (b) less powerful than
k-s or s-w)
1. seeking a good recent reference that generally discusses these
tests and their advantages
2 seeking a good recent reference that gives power of any of these
tests?
3. would like to know WHAT VALUE of such statistics woul UNDERMINE
NORMALITY ASSUMING TESTS , ANOVA, ANCOVA, REGRESSION.
this is not at all the same as asking whether distributions are
statsitically singificantly non-normal. texts assure us that ANOVA etc is
'robust to viloations of nomrality assumption' BUT how robust? how big
violations?
4. some normality tests are simply tests of symmetry since asymmetry
implies non-normality. BUT waht about tests of non-normality for
symmetric distributions, since high tails would still undermine standard
normal based tests?
many thanks for nay help that members of allstat can provide
diana kornbrot
*******************************************************************************
Dr. Diana Kornbrot
Reader in Mathematical Psychology
Associate Dean Research, Faculty of Health & Human Sciences
University of Hertfordshire
College Lane, Hatfield, Hertfordshire AL10 9AB, UK
voice: +44 0170 728 4626 fax: +44 0170 728 5073
email: [log in to unmask]
web: http://www.psy.herts.ac.uk/pub/D.E.Kornbrot/hmpage.html
*******************************************************************************
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|