Cross Validated is a question and answer site for people interested in statistics, machine learning, data analysis, data mining, and data visualization. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I want to know if this is a reasonable approach. I conducted the same experiment on two different species with small sample sizes (7-25 per species) in 4 years at 6 different sites. I measured damage to plants in two treatments.

There is unequal sampling across years and sites. Species, site and year differ in mean damage levels. The data are highly non-normal.

what I did:

  1. I got the residuals of a model damage= species +site(species) as a random effect + year
  2. Then I used a Wilcoxon rank sum test to test differences between treatments.

Does this seem reasonable?

share|improve this question
My first guess is no. In what way are your data "highly non-normal"? Are they counts, or binary classifications? If they're just skewed or something, bootstrapping may be a possibility. – gung Sep 3 '12 at 21:11
I'm confused as to what you tested with Wilcoxon - differences in residuals? If so, why? Wouldn't a nonlinear multilevel model answer the question more directly? – Peter Flom Sep 30 '12 at 12:43

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Browse other questions tagged or ask your own question.