DCA analysis results attached.
What does exactly length of gradients show us ?
The graphs collected for both PCA and CA analysis cases are attached. There is a problem with many zeros in the dataset too. Even if many rare species removed still graphs not readable. I just post them in case you can see any patterns which show which analysis should I follow. The many zeros and species is another problem I need to sort it out anyway. Is there any suggestions on that however?
A long gradient means very few species are shared by both ends of the gradient. You can add a constant to the data to account for the zeros or use Bray-Curtis dissimilarity. CA often produces more reliable ordinations than PCA in community ecology because it better models the non-linear responses of species to environmental and ecological gradients. But CA has the “arch” or “horseshoe” effect and compression of the ends of the gradient.
An alternative is to use the approach laid out by Legendre and Gallagher (2001) and developed in Legendre's Numerical Ecology book. Instead of assessing the gradient length you apply a Hellinger transformation to the data. Hellinger transformed data can be used in ordination techniques which rely on Euclidean distance, such as PCA, and addresses the double zero problem...
http://adn.biol.umontreal.ca/~numericalecology/Reprints/Legendre_&_Gallagher.pdf
Although there are also those who disagree with this approach...
https://eco.confex.com/eco/2010/techprogram/P24489.HTM
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Graphs