fixNonMonotHca: fixNonMonotHca
| fixNonMonotHca | R Documentation |
fixNonMonotHca
Description
Solving non-monotonous heights in HCA clustering.
Usage
fixNonMonotHca(hca, method = c("eps", "halfway"), eps = NULL)Arguments
hca |
an object of class |
method |
replacement method, one of 'eps', and
'halfway'.
The 'eps' method makes the non-monotonous heights to be monotonous
by shifting them just above the last monotonous height ( |
eps |
parameter of the 'eps' method, if |
Details
fixNonMonotHca replaces non-monotonous heights in a
dendrogram representing a hierarchical clustering with artificial
values making the heights monotonous.
However, please note that non-monotonous heights appear naturally in
dendrograms resulting from some HCA methods including
link{mhclust}, such that altering the heights result in
artificial dendrograms that do NOT represent the structure of the
clustered data. Please consider using HCA methods producing
monotonous before attempting to remove the non-monotonous heights,
e.g. for the purpose of graphical presentation of the dendrogram.
Value
An object of class link[stats]{hclust} having the
height component fixed.
Author(s)
Tomas Sieger, Karel Fiser
Examples
# simple example
d<-cbind(1:3,1:3)
hd<-mhclust(d)
# original dendrogram
print(hd$height)
hdFixed<-fixNonMonotHca(hd)
# dendrogram with resolved non-monotonous heights
print(hdFixed$height)
# another example
set.seed(1)
x<-cbind(runif(20),runif(20))
hx<-hclust(dist(x)^2,'cen')
hx1<-fixNonMonotHca(hx,method='eps')
hx2<-fixNonMonotHca(hx,method='halfway')
opar<-par(mfrow=c(1,3))
plot(hx,main='Original')
plot(hx1,main='Fixed by the "eps" method')
plot(hx2,main='Fixed by the "halfway" method')
par(opar)
