R package mdendro enables the calculation of agglomerative hierarchical clustering (AHC), extending the standard functionalities in several ways:
Native handling of both similarity and dissimilarity (distances) matrices.
Calculation of pair-group dendrograms and variable-group multidendrograms .
Implementation of the most common AHC methods in both weighted and unweighted forms: single linkage, complete linkage, average linkage (UPGMA and WPGMA), centroid (UPGMC and WPGMC), and Ward.
Implementation of two additional parametric families of methods: versatile linkage , and beta flexible. Versatile linkage leads naturally to the definition of two additional methods: harmonic linkage, and geometric linkage.
Calculation of the cophenetic (or ultrametric) matrix.
Calculation of five descriptors of the final dendrogram: cophenetic correlation coefficient, space distortion ratio, agglomerative coefficient, chaining coefficient, and tree balance.
Plots of the descriptors for the parametric methods.
All this functionality is obtained with two functions:
linkage may be considered as a replacement for functions
hclust (in package stats) and
agnes (in package cluster). To enhance usability and interoperability, the
linkage class includes several methods for plotting, summarizing information, and class conversion.
There exist two main ways to install mdendro:
Installation from CRAN (recommended method):
RStudio has a menu entry (Tools \(\rightarrow\) Install Packages) for this job.
Installation from GitHub (you may need to install first devtools):
install.packages("devtools") library(devtools) install_github("sergio-gomez/mdendro")
Let us start by using the
linkage function to calculate the complete linkage AHC of the
UScitiesD dataset, a matrix of distances between a few US cities:
library(mdendro) <- linkage(UScitiesD, method = "complete")lnk
Now we can plot the resulting dendrogram:
The summary of this dendrogram is:
## Call: ## linkage(prox = UScitiesD, ## type.prox = "distance", ## digits = 0, ## method = "complete", ## group = "variable") ## ## Binary dendrogram: TRUE ## ## Descriptive measures: ## cor sdr ac cc tb ## 0.8077859 1.0000000 0.7738478 0.3055556 0.9316262
In particular, you can recognize the calculated descriptors:
cor: cophenetic correlation coefficient
sdr: space distortion ratio
ac: agglomerative coefficient
cc: chaining coefficient
tb: tree balance
It is possible to work with similarity data without having to convert them to distances, provided they are in range [0.0, 1.0]. A typical example would be a matrix of non-negative correlations:
<- as.dist(Harman23.cor$cov) sim <- linkage(sim, type.prox = "sim") lnk plot(lnk, main = "Harman23")