Means question: LSD means outside of data range

[astat]

I just ran PROC GLM on a data set with a value range of 0.0278 to 0.1668, and the means shown in the LSD tables ranged from 0.861 to 1.06. I can understand some small differences in means calculations, but how do the LSD means end up completely out of range of my original data set? The LSD means also ranked my treatments in a completely different order than the arithmetic means. I thought maybe I was being too ambitious with "by groups" processing, so I ran a simple PROC BOXPLOT to check, and the box plots came out skewed, too, with the upper whiskers all reaching 4 or 5!
:headbang:

Any ideas on what I might have done to offend the SAS gods?
Thanks! Message was edited by: astat

[statsplank]

Hi,

It seems what you violated is a normality assumption. You may want to apply some transformation to your response to make it approximately normal.