CONJPM@UCSCVM.BITNET (Jim Mulherin) (02/06/90)
I have a user who is looking for a way in SAS to do non-parametric multiple comparison among more than two means (4 right now). He has looked in Sokal and Rohlf (he's a biology grad student) and SAS (which says to use Proc Rank followed by Proc GLM or ANOVA). The problem is getting spurious significance in doing lots of tests. He says he wants to do the equivalent of a Tukey-Kramer multiple range test. What do you suggest? He has SAS 5.18 on CMS and MVS and SAS 6.03 on PCs available. Data is Elephant Seal rates of behaviors (esp. male threat). A large portion (20-80%) are zeros (no threat behavior observed). This creates a large number of tied ranks.