Hierarchical Cluster Analysis of Spectra in R

Example of hierarchical cluster analysis on spectra. First the distances are calculated for spectra matrix by pearson.dist function, then dendrogram is created with hclust function (here Ward’s algorithm is selected as linkage method).

# dendrogram
dist <- pearson.dist(test [[]])
dendrogram <- hclust(dist , method = "ward")
test$clusters <- as.factor (cutree ( dendrogram, k=5 ))
# mean spectra for clusters
cluster.means <- aggregate ( test, test$clusters, mean)
cols <- c ("blue", "cyan","green" , "yellow" ,"red" )
plot(cluster.means, stacked = ".aggregate" , fill = ".aggregate", col = cols, wl.reverse= TRUE, xoffset=150,lines.args = list( lwd = 3, lty = 1 ), wl.range =c( 800~2250 ,2400 ~3800))
#clusters mapping
levels ( test$clusters ) <- c (1,2 ,3 , 4 ,5 )
plotmap(test,    z  =  test$clusters)
# dendrogram with colour bars
plot(dendrogram )
plot(dendrogram, labels = FALSE, hang=0 )
points ( seq_along (dendrogram $order) , rep (-3, length (dendrogram$order)), col= cols [test$ clusters [dendrogram $order]] , pch ="|")
# plus extra rectangles
rect.hclust(dendrogram, k =5, border ="red")