I am using Redundancy analysis to explore the effect of plant community composition on invertebrate abundances (log-transformed values) within sampling sites. I am using the vegan package in R and have been getting a result where my plant community explains more than 60% of the variation in the invertebrate community, but is not significant (through anova.cca function). When I use an environmental matrix to explain the invertebrates, it explains much less variation but is usually significant. I'm wondering that since my plant community is based on presence/absence, if this is what is causing the issue. I would appreciate any suggestions or direction to help with this.
Hi Kelllie,
Strange one. Only thing I can think of is the anova is mistaking your binary plants data for real 0's and 1's. I've never done an RDA with this kind of data as a predictor, but you could try re-coding as a single vector of factors and see what comes out then. Then you've got a gradient of sorts of plant presence rather than a series of dummy variables which makes more sense to me for RDA. Something like this:
#rng inverts data
inverts
Thanks for the suggestion Malcolm. I'll give it a try and see what happens!
Hi Kellie,
I would transform the plant community matrix into a set of 'coordinates' using either PCA or nMDS, and giving you (a minimum) of 2 variables which explain variation within the plant community. If the data is presence/absence, you will need to do the relevant transformation before you run it in PCA, however the matrix can be used as is in nMDS.
You may want to consider variation partitioning as well, that will allow you to test for the independent effects of plant community and environment on the invertebrates, as well as the combined effect.
The vegan package in R covers this quite well!
Cheers
Chris
- Badges
- Science topic
- Similar topics
- Biological Science
- Zoology
- Invertebrate Zoology