consensusRDA: Calculates a consensus RDA

In ordiconsensus: Consensus of canonical ordinations through the canonical redundancy analysis

Description

Calculates a consensus RDA. That is, a consensus for a series of RDA performed on the same data using different association coefficients.

Usage

consensusRDA(ordires, ordisigniaxis, resp.var, expl.var, pval = 0.05, scaling = 2)

Arguments

ordires	A list of rda or capscale result object that includes a series of RDA or distance-based RDA (db-RDA) performed with different association coefficients on the same data.
ordisigniaxis	A list of anova.cca object where each axis of each RDA (or db-RDA) given in ordires was tested for significance. This argument can also be a vector defining the number of significant axes in each RDA. See details.
resp.var	A matrix of the response variables (species) used to construct all the RDAs (or db-RDAs) in ordires.
expl.var	A matrix of the explanatory variables used to construct all the RDAs (or db-RDAs) in ordires.
pval	Numeric. P-value threshold to select the number of axes to use. This argument is only active if a list of anova.cca object is given for the argument ordisigniaxis, otherwise it is not considered. Default is 0.05.
scaling	Type of scaling used to project the consensus RDA result. A distance scaling (scaling = 1) or a correlation scaling (scaling = 2) can be used. Default is scaling = 1. See details for more information.

Details

For the argument ordisigniaxis, if a vector defining the number of significant axes is given, it is assumed that the significant axes are selected in sequential order from the first axis. Although it is possible for the scaling to be 3 (it is available in the vegan package), this scaling should only be used for canonical correspondence analysis (CCA); it does not make any sense to use in the RDA framework.

Value

value	A vector of eigenvalues associated to the axes of the consensus RDA
siteConsensus	A matrix of consensus sites scores
spConsensus	A matrix of consensus species scores
descConsensus	A matrix of consensus canonical coefficients

Author(s)

F. Guillaume Blanchet

Examples

###########################################################################
### This example reproduces Figure 7c of Blanchet et al. (in press)
### 
### However, for illustration purposes RDA axes were tested using anova.cca
### This only has minor influence on the triplot. 
###########################################################################
data(beetle)
data(beetle.expl)

### Construct results object
ndis<-9
ordiRes<-vector("list",length=ndis)

### RDA species profile
sp<-beetle/apply(beetle,1,sum)
ordiRes[[1]]<-rda(sp~.,data=beetle.expl)
	
### RDA chord
chord<-beetle/sqrt(apply(beetle^2,1,sum))
ordiRes[[2]]<-rda(chord~.,data=beetle.expl)
	
### RDA Hellinger
hell<-decostand(beetle,method="hellinger")
ordiRes[[3]]<-rda(hell~.,data=beetle.expl)

### db-RDA Bray-Curtis
bray<-sqrt(vegdist(beetle,method="bray"))
ordiRes[[4]]<-capscale(bray~.,data=beetle.expl,comm=beetle)
	
### db-RDA square-root Bray-Curtis
bray.sqrt<-sqrt(vegdist(beetle^0.5,method="bray"))
ordiRes[[5]]<-capscale(bray.sqrt~.,data=beetle.expl,comm=beetle^0.5)
	
### db-RDA fourth-root Bray-Curtis
bray.fort<-sqrt(vegdist(beetle^0.25,method="bray"))
ordiRes[[6]]<-capscale(bray.fort~.,data=beetle.expl,comm=beetle^0.25)
	
### db-RDA modified Gower log 2
beetleLog2<-decostand(beetle, "log",logbase=2)
mGowerLog2<-vegdist(beetleLog2, "altGower")
ordiRes[[7]]<-capscale(mGowerLog2~.,data=beetle.expl,comm=beetleLog2)
	
### db-RDA modified Gower log 5
beetleLog5<-decostand(beetle, "log",logbase=5)
mGowerLog5<-vegdist(beetleLog5, "altGower")
ordiRes[[8]]<-capscale(mGowerLog5~.,data=beetle.expl,comm=beetleLog5)
	
### db-RDA modified Gower log 10
beetleLog10<-decostand(beetle, "log",logbase=10)
mGowerLog10<-vegdist(beetleLog10, "altGower")
ordiRes[[9]]<-capscale(mGowerLog10~.,data=beetle.expl,comm=beetleLog10)

#----------------
### Test RDA axis 
#----------------
ordiResTest<-vector("list",length=ndis)

## Not run: 
for(i in 1:ndis){
	ordiResTest[[i]]<-anova.cca(ordiRes[[i]],by="axis",cutoff=0.1,perm.max=100)
}

## End(Not run)

ordiResTest<-c(3,3,4,4,4,4,3,4,4) ## This output is an alternative to the previous code


### Consensus RDA
consRDA<-consensusRDA(ordiRes,ordiResTest,beetle,beetle.expl)
summary(consRDA)
XaxisLabels<-paste("Axis 1 - ",round(consRDA$values[1]/sum(consRDA$values),4)*100,sep="")
YaxisLabels<-paste("Axis 2 - ",round(consRDA$values[2]/sum(consRDA$values),4)*100,sep="")

plot(consRDA$siteConsensus[,1:2],pch=19,xlab=XaxisLabels,ylab=YaxisLabels,las=1)
abline(h=0,v=0,lty=3)

arrows(0,0,consRDA$spConsensus[,1]*10,consRDA$spConsensus[,2]*10,
       col="red",length=0.1,angle=12)
text(consRDA$spConsensus[,1:2]*10,labels=rownames(consRDA$spConsensus),col="red")

arrows(0,0,consRDA$descConsensus[,1]*0.4,consRDA$descConsensus[,2]*0.4,
      col="blue",length=0.1,angle=12)
text(consRDA$descConsensus[,1:2]*0.4,labels=rownames(consRDA$descConsensus),col="blue")

ordiconsensus documentation built on May 2, 2019, 4:38 p.m.