0

I am trying to plot an RDA correlation triplot (using plot(rda_model)) using the vegan package in R, however I am confused about which 'scaling' to use.

In Zuur (2007, Analysing Ecological Data), scaling refers to the power of the diagonal matrix containing the square root of the eigenvalues and states that for the correlation triplot scaling=2. However, Jari Oksanen (Design decisions and implementation details in vegan) uses this term (when using plot(rda)) for scaling either sites or species by the eigenvalues.

I would really appreiate advice on how to do a correlation triplot- whichever thing scaling refers to.

M1=rda(SPECr~.,data = CHEMr,scale=TRUE)

summary(M1)

plot(M1,main="Correlation triplot",cex=1.1,scaling=2)#is this correct?

var.scall=scores(M1,choices=1:2,display="sp",scaling=2)

arrows(0,0,var.scall[,1],var.scall[,2],length=0,lty=1,col="red")
Alex93
  • 1
  • 1
  • He's Zuur (with two 'u's and one 'r'). The meaning of the 'scaling' parameter in vegan is similar as in CANOCO 3 and in ter Braak's articles. Somehow I think that Zuur's description is similar to CANOCO & vegan and different from your description, but I haven't seen the book (too expensive). – Jari Oksanen Sep 08 '19 at 16:38
  • Thanks for the advice (and for pointing out the typo)! – Alex93 Sep 08 '19 at 19:10
  • In recent vegan, you can choose correlation-based scores for the plot you want by specifying `correlation = TRUE` with `scaling = “sites”`, or `”species”` or “`symmetric”` depending on what you want to focus on. IIRC, the correlation-based score were negative values in Canoco-speak. In CCA you use `hill = TRUE` rather than `correlation = TRUE`. – Gavin Simpson Sep 08 '19 at 20:02

0 Answers0