[R-sig-eco] biplot with most important species

[syrovat]

Hi Bjorn,

A common practice in simplifying ordinatin plots is to keep species that 
show the best fit into the ordination space.
The position of such species in the ordination diagram really (may) tell 
something.
In linear methods like PCA a linear fit is suitable. This can be 
computed using vegan::envfit function.

For this I have a very simple function reduce.pca, which you can source 
using this line of code:
source(url("http://vitsyrovatka.info/lib/exe/fetch.php?media=zpradat:functions:reduce.pca.r"))

The code and example may be viewed here (skip the Czech introduction):
http://vitsyrovatka.info/doku.php?id=zpradat:cs:fun:reducepca

Your criterion on species selection is more complicated. You can change 
the function accordingly.
I am not completely sure but a solution might be multiplying species fit 
by species variance,
then ordering species acoording to this fit*variance and add species 
until the desired variance is achieved.
I might be wrong, anybody feel free to correct me, please.

Cheers,
Vit


Dne 2016-10-12 12:44, Bjorn napsal:
> Hi all,
> 
> there might be a simple solution to do this, but I don't seem to
> manage to find it. I have done a PCA using vegan. I now want to create
> a biplot with only those species that explain a certain (cumulative)
> amount of the variation along the first 2 PC axes (in order to keep
> things clear). How could this be done?
> 
> Thanks in advance!
> 
> Kind regards,
> 
> Bjorn
> 
> _______________________________________________
> R-sig-ecology mailing list
> R-sig-ecology at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-ecology