I need some advice on how to proceed with my data analysis. I have 3 groups (Archaea, Bacteria and Eukarya). Each group has unequal number of individual species (70, 651, and 244 respectively). Each specie was given 7 different treatments (A, B, C, D, E, F, and G) and I counted the number of domains (you can call it 'y').

Now the problems are: Counts for treatment A are extremely high compared to counts of other treatments. Treatment G has the lowest counts. Other treatments have more or less similar counts. Data is non-normal, does not have homogenous variances, unequally replicated and any help would be appreciated.

  • Which test should I use?
  • Any transformation?
  • $\begingroup$ Please tell us what you are trying to find out. What is your research question(s)? What hypotheses? $\endgroup$ – Peter Flom Aug 14 '11 at 12:40

Some of your questions about how to respond to violations of anova assumptions may be answered by this flow chart .

However, more important may be that you look into multi-level modeling, a.k.a. hierarchical linear modeling or nested modeling. This is because you have species nested within groups.


One option is to use a permutation test. If you are testing whether the treatments have an effect then you could permute observations within each species and compare the resulting statistic of interest from the various permutations to the original data. Note that this is testing that within each species the distribution (shape, mean, and variance) is identical between the treatments, but could differ between species. If your hypothesis is different than this (as Peter mentioned we need details to be of more help) then you would need a different test.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.