-1

I am using R with the vegan package to analyse the significance of environmental factors on bacterial communities (OTU table) at different sites. However I got a few issues which I am not sure how to address, and if my approach is suitable. Would be awesome to get some feedback.

1) I have one table with 15 sites with 380 species and a table with 8 environmental factors measured at each site. Is that too small for cca? Any suggestions for a better method?

2) I have used cca(), which gave me very similar Eigenvalues for Constrained/Unconstrained- I was under the impression that the constrained Eigenvalues need to be lower to be a suitable analysis- might be problem from small dataset?

3) Some of my environmental factors are correlated, I checked their VIF and I had values between 1 and 61. Apparently <10 is acceptable- best to remove high value factors?

Hope this makes, sense, thank you very much in advance!

Cheers Mel

Melanie Br
  • 1

1 Answers1

0
  1. 15 observations is not much for significance tests.
  2. If eigenvalues are very similar in constrained and unconstrained analysis, you probably have too many constraints. This comes back to small data sets and large number of constraining variables.
  3. Not necessarily: high VIF means that the set of your variables is autocorrelated, but it may not be the best choice to select the one with highest VIF: after you reduce your set, a removed variable may have low VIF and be perfectly OK. You should build a small model (not many variables) and then check that it is also acceptable for VIFs.
Jari Oksanen
  • 3,287
  • 1
  • 11
  • 15