I am doing a cluster analysis and I want to count the number of occurences of a certain variable in a leaf of a pruned tree. Below is a simplified example where the pruned tree has only three branches. I now want to know the number of As and Bs in the three differnt branches/leafs. How can I get those?
rm(list=ls(all=TRUE))
mylabels <- matrix(nrow=1, ncol = 20)
mylabels[1,1:10] <- ("A")
mylabels[1,11:20] <- ("B")
myclusterdata <- matrix(rexp(100, rate=.1), ncol=100, nrow=20)
rownames(myclusterdata)<-mylabels
hc <- hclust(dist(myclusterdata), "ave")
memb <- cutree(hc, k = 3)
cent <- NULL
for(k in 1:3){
cent <- rbind(cent, colMeans(myclusterdata[memb == k, , drop = FALSE]))
}
hc1 <- hclust(dist(cent)^2, method = "cen", members = table(memb))
# whole tree
plot(as.dendrogram(hc),horiz=T)
# pruned tree (only 3 branches)
plot(as.dendrogram(hc1),horiz=T)
OK I figured it out. The elements of the leaf are in memb... So rearranging them and combining it provides the results. Below is the code for the example
rm(list=ls(all=TRUE))
mylabels <- matrix(nrow=1, ncol = 20)
mylabels[1,1:10] <- ("A")
mylabels[1,11:20] <- ("B")
myclusterdata <- matrix(rexp(100, rate=.1), ncol=100, nrow=20)
rownames(myclusterdata)<-mylabels
hc <- hclust(dist(myclusterdata), "ave")
memb <- cutree(hc, k = 3)
cent <- NULL
for(k in 1:3){
cent <- rbind(cent, colMeans(myclusterdata[memb == k, , drop = FALSE]))
}
hc1 <- hclust(dist(cent)^2, method = "cen", members = table(memb))
# whole tree
plot(as.dendrogram(hc),horiz=T)
# pruned tree (only 3 branches)
plot(as.dendrogram(hc1),horiz=T)
# identify the percentages of A and B
var_of_interest <- levels(as.factor(names(memb)))
leaf_number <- levels(as.factor(memb))
counter <- matrix(nrow=length(leaf_number), ncol = length(var_of_interest))
for (i in seq(1:length(leaf_number))) {
for (j in seq(1:length(var_of_interest))) {
counter[i,j] <- length(memb[names(memb)==var_of_interest[j] & memb==leaf_number[i]])
}
}
counter[,2]/(counter[,1]+counter[,2])