[R-sig-eco] PCA with vegan

[amelie_can]

Thank you Jari for your help! A few precision follows....

My project evaluates temporal dynamics of vegetation in a restoration
project. So, there is tons of double zero and indeed, a few species are
highly dominant in the data. Actually, I should have precised that my matrix
of species abundance have quite a "normal variation" of species distribution
through space with values around 34 and 24% for the first two axis. So,
Hellinger transformation seems to be appropriate here. 

The 84 % is the result obtained with the matrix of community mean weight
analysis, which give the weight of ecological characteristics of the plant
present. That matrix was not Hellinger transformed but we calculated species
abundances based on their relative contribution of each species to its
community (pi) to ensure comparability of two different dataset. This is a
kind of standardization. 

My supervisor said a PCA analysis must be either standardized with either a
correlation or a covariance technique before hand. A correlation seems more
appropriate for my data, which is what scale = TRUE does, right? But I agree
with you that species characteristics should not contribute the same amount
to the variance as the main hypothesis I am trying to answer is to find
which characteristics of the species explains their presence in the site
studied. 

I tried to rerun the analysis with Hellinger transformed data and obtained
80% of my first axis, so it does not seems appropriate either.

This is the code I used: pca.results= rda(x, scale=FALSE). This analysis
seems right to me for the answer I am trying to get.  

I inspected the variance as you proposed and it seems that the first 8
variables are the most important. 
variable 1: 0.0527226189 
variable 2: 0.0452730816 
variable 3: 0.0432572385 
variable 4: 0.0347574432 
variable 5: 0.0295173212 
variable 6: 0.0257054085 
variable 7: 0.0156271508 
variable 8: 0.0105172472 
variable 9: 0.0085778112 
variable 10: 0.0047607918 
variable 11: 0.0042802434 
variable 12: 0.0016321132 
variable 13: 0.0007667426 

I guess like you said it means that my traits of the species present did not
vary that much and were mainly explained by the first 8. If you see any
mistake in this reasoning, please let me know! 


--
View this message in context: http://r-sig-ecology.471788.n2.nabble.com/PCA-with-vegan-tp6446676p6450406.html
Sent from the r-sig-ecology mailing list archive at Nabble.com.