====== Colored dendrograms ====== ===== A2R ===== Easter (April 7) 2012 On the **R-GraphGallery** we can find the [[http://addictedtor.free.fr/graphiques/RGraphGallery.php?graph=79|Colored Dendrogram]] example. - using R install the R-package ''fpc'' ([[http://cran.r-project.org/web/packages/fpc/index.html|CRAN]]) - download the R-package ''A2R'' ([[http://addictedtor.free.fr/packages/|CRAN]]) on the file ''A2R_0.0-4.tar.gz''. There are some problems installing ''A2R'' ([[https://stat.ethz.ch/pipermail/r-help/2008-May/161470.html|R-help]],[[http://stackoverflow.com/questions/6003680/r-colored-dendrogram-suggestions |Stackoverflow]]). We proceed as follows: - download [[http://cran.r-project.org/bin/windows/Rtools/|rtools]] and install them; - include ''rtools'' into ''PATH'':\\ ''c:\Rtools\bin;c:\Rtools\gcc-4.6.3\bin;c:\MiKTeX\miktex\bin;C:\R\R-2.14.0\bin\i386;c:\windows;c:\windows\system32;'' - download [[http://www.activestate.com/activeperl/downloads|activeperl]] and install it; - in the Run window run the command:\\ ''rcmd INSTALL e:/zip/R/A2R_0.0-4.tar.gz'' Now the request ''library(A2R)'' in R should work. See also [[http://stackoverflow.com/questions/6009423/a2r-library-colored-dendrogram-allow-more-than-six-chars-per-label|More than six chars per label]]. [[http://addictedtor.free.fr/packages/A2R/lastVersion/html/A2Rplot.hclust.html|A2R docs]] It seems that it works also by simply downloading the content of the [[http://addictedtor.free.fr/packages/A2R/lastVersion/R/|last version/R]] into the map A2R and then


> setwd("C:/Users/Batagelj/work/R/A2R")
> require(fpc)
> require(grid)
> source("C:\\Users\\Batagelj\\work\\R\\A2R\\A2R.R")

> # examples with state.x77
> d77 <- dist(state.x77)
> h77 <- hclust(d77)
> A2Rplot(h77, k=4, knot.pos="mean", type="tri")

{{notes:pics:x77dendro.png}} ===== Coloring leaves of dendrogram ===== Searching for solution how to install ''A2R'' I found an alternative solution how to [[http://www.opensubscriber.com/message/r-help@stat.math.ethz.ch/3554569.html|Coloring leaves in a hclust or dendrogram plot]]. I slightly modified the proposed solution so that the colors and sizes can be specified by 5 tables: * labn[i] - node label * labc[i] - label color * labs[i] - label size * dotc[i] - dot color * dots[i] - dot size The node with the label labn[i] will get the attributes specified in labc[i], labs[i], dotc[i] and dots[i].


> colLab <- function(n){ 
+  if(is.leaf(n)){ 
+    a <- attributes(n) 
+    inds <- which(labn == a$label)
+    if ( length(inds) == 1 ){ i <- inds[1]
+      attr(n, "nodePar") <- c(a$nodePar, list(lab.col = labc[i], lab.cex=labs[i], 
+           col=dotc[i], cex=dots[i], pch=16 )) 
+    } else { 
+ #     attr(n, "nodePar") <- c(a$nodePar, list(lab.col = "red", lab.cex=.7, 
+ #          col="red", cex=pop[n], pch=16)) 
+    }} 
+  n 
+  } 
> 

> K <- sample(50)/50; B <- sample(50)/50
> y <- data.frame(K,B)
> labn <- paste('T',1:50,sep='')
> head(labn)
[1] "T1" "T2" "T3" "T4" "T5" "T6"
> rownames(y) <- labn
> head(y)
      K    B
T1 0.50 0.58
T2 0.92 0.80
T3 0.22 0.10
T4 0.34 0.94
T5 0.72 0.74
T6 0.96 0.62
> hc <- hclust(dist(y), "ward")
> plot(hc)
> cla <- (y$K>0.5)+2*(y$B>0.5)+1
> head(cla)
[1] 3 4 1 3 4 4
> col <- rep("black",50)
> col[cla==1] <- "red"
> col[cla==2] <- "blue"
> col[cla==3] <- "green"
> col[cla==4] <- "magenta"
> head(col)
[1] "green"   "magenta" "red"     "green"   "magenta" "magenta"
> labc <- col; dotc <- col
> labs <- rep(0.5,50); dots <- rep(0.3,50)
> plot(y,col=col,pch=16)
> text(y$K,y$B-0.016,1:50,cex=0.5)

> dend <- as.dendrogram(hc,hang=-1)
> dend_colored <- dendrapply(dend, colLab)
> plot(dend_colored)


> clu <- cutree(hc,k=4)
> head(clu)
T1 T2 T3 T4 T5 T6 
 1  1  2  3  1  4 
> col[clu==1] <- "red"
> col[clu==2] <- "blue"
> col[clu==3] <- "green"
> col[clu==4] <- "magenta"
> head(col)
[1] "red"     "red"     "blue"    "green"   "red"     "magenta"
> labc <- col; dotc <- col
> dend <- as.dendrogram(hc,hang=-1)
> dend_colored <- dendrapply(dend, colLab)
> plot(dend_colored)