1

我正在做一个聚类分析，我想计算某个变量在修剪树的叶子中出现的次数。下面是一个简化的示例，其中修剪的树只有三个分支。我现在想知道三个不同分支/叶子中的 As 和 B 的数量。我怎样才能得到那些？

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)
4

1 回答 1

0

好的，我想通了。叶子的元素在 memb 中......所以重新排列它们并组合它提供了结果。下面是示例的代码

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])
于 2015-11-11T06:52:55.637 回答