consensus {agricolae}

consensus of clusters

Description

The criterion of the consensus is to produce many trees by means of boostrap and to such calculate the relative frequency with members of the clusters.

Usage

consensus(data,distance=c("binary","euclidean","maximum","manhattan",
"canberra", "minkowski", "gower","chisq"),method=c("complete","ward","single","average",
"mcquitty","median", "centroid"),nboot=500,duplicate=TRUE,cex.text=1, 
col.text="red", ...)

Arguments

data

data frame

distance

method distance, see dist()

method

method cluster, see hclust()

nboot

The number of bootstrap samples desired.

duplicate

control is TRUE other case is FALSE

cex.text

size text on percentage consensus

col.text

color text on percentage consensus

...

parameters of the plot dendrogram

Details

distance: "euclidean", "maximum", "manhattan", "canberra", "binary", "minkowski", "gower", "chisq". Method: "ward", "single", "complete", "average", "mcquitty", "median", "centroid". see functions: dist(), hclust() and daisy() of cluster.

Value

table.dend

The groups and consensus percentage

dendrogram

The class object is hclust, dendrogram plot

duplicate

Homonymous elements

Author(s)

F. de Mendiburu

References

An Introduction to the Boostrap. Bradley Efron and Robert J. Tibshirani. 1993. Chapman and Hall/CRC

See Also

hclust, hgroups, hcut

Examples

library(agricolae)
data(pamCIP)
# only code
rownames(pamCIP)<-substr(rownames(pamCIP),1,6)
output<-consensus( pamCIP,distance="binary", method="complete",nboot=5)
# Order consensus
Groups<-output$table.dend[,c(6,5)]
Groups<-Groups[order(Groups[,2],decreasing=TRUE),]
print(Groups)
## Identification of the codes with the numbers.
cbind(output$dendrogram$labels)
## To reproduce dendrogram
dend<-output$dendrogram
data<-output$table.dend
plot(dend)
text(data[,3],data[,4],data[,5])
# Other examples
# classical dendrogram
dend<-as.dendrogram(output$dendrogram)
plot(dend,type="r",edgePar = list(lty=1:2, col=2:1))
text(data[,3],data[,4],data[,5],col="blue",cex=1)
plot(dend,type="t",edgePar = list(lty=1:2, col=2:1))
text(data[,3],data[,4],data[,5],col="blue",cex=1)
## Without the control of duplicates
output<-consensus( pamCIP,duplicate=FALSE,nboot=5)
## using distance gower, require cluster package.
# output<-consensus( pamCIP,distance="gower", method="complete",nboot=5)