Subsets of a dataset as separate dendrograms, but in the same plot

问题

I know I can plot a dendrogram as follows

library(cluster)
d <- mtcars
d[,8:11] <- lapply(d[,8:11], as.factor)

gdist <- daisy(d, metric = c("gower"), stand = FALSE)
dendro <- hclust(gdist, method = "average")
plot(as.dendrogram(dendro))

However I have some groups identified (eg. by an iterative classification method), given as the last column in d

G <- c(1,2,3,3,4,4,5,5,5,5,1,2,1,1,2,4,1,3,4,5,1,7,4,3,3,2,1,1,1,3,5,6)
d$Group <- G

head(d)
                   mpg cyl disp  hp drat    wt  qsec vs am gear carb  Group
Mazda RX4         21.0   6  160 110 3.90 2.620 16.46  0  1    4    4     1
Mazda RX4 Wag     21.0   6  160 110 3.90 2.875 17.02  0  1    4    4     2
Datsun 710        22.8   4  108  93 3.85 2.320 18.61  1  1    4    1     3
Hornet 4 Drive    21.4   6  258 110 3.08 3.215 19.44  1  0    3    1     3
Hornet Sportabout 18.7   8  360 175 3.15 3.440 17.02  0  0    3    2     4
Valiant           18.1   6  225 105 2.76 3.460 20.22  1  0    3    1     4

I am trying to plot all the dendrograms together on the same plot with the same scale. The groups with only a single member also needs to be plotted. (group 6 and 7)

I am able to plot individual dendrograms for subset of the data except when number of members in a group is only one. But I don't think this is the right approach.

layout(matrix(1:9, 3,3,byrow=TRUE))

gdist <- as.matrix(gdist)

N <- max(G)
for (i in 1:N){
  rc_tokeep <- row.names(subset(d, G==i))
  dis <- as.dist(gdist[rc_tokeep, rc_tokeep])
  dend <- hclust(dis, method = "average")
  plot(as.dendrogram(dend))
}

The loop is giving this error for the last two groups. (6 and 7) having only a single member.

Error in hclust(dis, method = "average") : 
  must have n >= 2 objects to cluster

Essentially I wan't to reproduce these type of plots. The clusters with single members are also plotted here.

回答1:

If you want to mimic the last few graphs, you can do something like this:

N <- max(G)
layout(matrix(c(0,1:N,0),nc=1))

gdist <- as.matrix(gdist)

for (i in 1:N){
    par(mar=c(0,3,0,7))
    rc_tokeep <- row.names(subset(d, G==i))
    if(length(rc_tokeep)>2){ #The idea is to catch the groups with one single element to plot them differently
        dis <- as.dist(gdist[rc_tokeep, rc_tokeep])
        dend <- hclust(dis, method = "average")
        plot(as.dendrogram(dend),horiz=TRUE,
                 xlim=c(.8,0),axes=FALSE) # giving the same xlim will scale all of them, here i used 0.8 to fit your data but you can change it to whatever
        }else{
            plot(NA,xlim=c(.8,0),ylim=c(0,1),axes=F,ann=F)
            segments(0,.5,.1,.5) #I don't know how you intend to compute the length of the branch in a group of 1 element, you might want to change that
            text(0,.5, pos=4,rc_tokeep,xpd=TRUE)
            }
}

With your example it gives:

If you want to add the scale you can add a grid in all graphs and a scale in the last one:

N <- max(G)
layout(matrix(c(0,1:N,0),nc=1))

gdist <- as.matrix(gdist)

for (i in 1:N){
    par(mar=c(0,3,0,7))
    rc_tokeep <- row.names(subset(d, G==i))
    if(length(rc_tokeep)>2){
        dis <- as.dist(gdist[rc_tokeep, rc_tokeep])
        dend <- hclust(dis, method = "average")
        plot(as.dendrogram(dend),horiz=TRUE,xlim=c(.8,0),xaxt="n",yaxt="n")
        abline(v=seq(0,.8,.1),lty=3) #Here the grid
        }else{
            plot(NA,xlim=c(.8,0),ylim=c(0,1),axes=F,ann=F)
            segments(0,.5,.1,.5)
            text(0,.5, pos=4,rc_tokeep,xpd=TRUE)
            abline(v=seq(0,.8,.1),lty=3) #Here the grid
            }
    }
axis(1,at=seq(0,.8,.1)) #Here the axis

And finally if you want to even the spaces between the different branches in the resulting plot, you can use table(d$Group) to get the number of members of each group and use it as a height for each subplot:

N <- max(G)

layout(matrix(c(0,1:7,0),nc=1), height=c(3,table(d$Group),3)) #Plus the height of the empty spaces.

gdist <- as.matrix(gdist)

for (i in 1:N){
    par(mar=c(0,3,0,7))
    rc_tokeep <- row.names(subset(d, G==i))
    if(length(rc_tokeep)>2){
        dis <- as.dist(gdist[rc_tokeep, rc_tokeep])
        dend <- hclust(dis, method = "average")
        plot(as.dendrogram(dend),horiz=TRUE,xlim=c(.8,0),xaxt="n",yaxt="n")
        abline(v=seq(0,.8,.1),lty=3)
        }else{
            plot(NA,xlim=c(.8,0),ylim=c(0,1),axes=F,ann=F)
            segments(0,.5,.1,.5)
            text(0,.5, pos=4,rc_tokeep,xpd=TRUE)
            abline(v=seq(0,.8,.1),lty=3)
            }
    }
axis(1,at=seq(0,.8,.1))

来源：https://stackoverflow.com/questions/24284928/subsets-of-a-dataset-as-separate-dendrograms-but-in-the-same-plot

标签

plot

cluster-analysis

dendrogram

hclust