Hi All, How reliable is Dunnett's comparison (w/control) when applied to ranked data ? This appears to be an ad hoc nonparametric multiple comparison with a test statistic based on group rank sums in place of group means and a standard error of sqrt[n(np)(np+1))/6]. My guess is that it's unstable for small samples.
I found reference to it in Zar, "Biostatistical Analysis" (1974) (only edition in our library). The standard error is cited to Wilcoxon and Wilcox (1964)(rare ?). Why 6 instead of 12 ? thanks, Al Barron Metuchen, NJ __________________________________________________ Do You Yahoo!? Yahoo! Sports - sign up for Fantasy Baseball http://sports.yahoo.com ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at http://jse.stat.ncsu.edu/ =================================================================
