R/pcoatest.R

In dave: Functions for "Data Analysis in Vegetation Ecology"

pcoatest <-
function(veg,y=1) {
    sveg<- veg^y
    #
    # coordinates of all 6 methods
    #
    vdm<- vegdist(sveg,method="euclidean")
    oute<- pco(vdm)
    
    vdm<- vegdist(sveg,method="manhattan")
    outm<- pco(vdm)
    
    # chord distance
    vdm<- dist(decostand(sveg,"norm"),method="euclidean")
    outco<- pco(vdm)
    
    vdm<- vegdist(sveg,method="canberra")
    outc<- pco(vdm)
    
    vdm<- vegdist(sveg,method="bray")
    outb<- pco(vdm)
    
    vdm<- as.dist((1 - cor(t(sveg)))/2)
    outcor<- pco(vdm)
    # now clustering
    o.hclr<- hclust(vdm,method="ward.D2")
    o.grel<- cutree(o.hclr,k=5)
    
    
    # PCA
    outpca<- prcomp(sveg)
    pcscor<- scores(outpca,display="sites",choices=1:2)
    #
    # procrustes analysis for all 6 methods
    #
    # euclid
    proce<- procrustes(pcscor,scores(oute))
    
    # manhattan
    procm<- procrustes(pcscor,scores(outm))
    
    # chord
    procco<- procrustes(pcscor,scores(outco))
    
    # canberra
    procca<- procrustes(pcscor,scores(outc))
    
    # bray
    procb<- procrustes(pcscor,scores(outb))
    
    # 1-correlation
    procc<- procrustes(pcscor,scores(outcor))
    #
    # output list
    #
    pcovar<- list(nrel=nrow(sveg),nspec=ncol(sveg),y=y,euclidpca=proce$Yrot,euclidpco=proce$X,manhpca=procm$Yrot,manhpco=procm$X,cordpca=procco$Yrot,cordpco=procco$X,canpca=procca$Yrot,canpco=procca$X,bpca=procb$Yrot,bpco=procb$X,corpca=procc$Yrot,corpco=procc$X,grel=o.grel)
}