cl_ultrametric: Ultrametrics of Hierarchies

View source: R/ultrametric.R

cl_ultrametricR Documentation

Ultrametrics of Hierarchies

Description

Compute the ultrametric distances for objects representing (total indexed) hierarchies.

Usage

cl_ultrametric(x, size = NULL, labels = NULL)
as.cl_ultrametric(x)

Arguments

x

an R object representing a (total indexed) hierarchy of objects.

size

an integer giving the number of objects in the hierarchy.

labels

a character vector giving the names of the objects in the hierarchy.

Details

If x is not an ultrametric or a hierarchy with an ultrametric representation, cl_ultrametric uses cophenetic to obtain the ultrametric (also known as cophenetic) distances from the hierarchy, which in turn by default calls the S3 generic as.hclust on the hierarchy. Support for a class which represents hierarchies can thus be added by providing as.hclust methods for this class. In R 2.1.0 or better, cophenetic is an S3 generic as well, and one can also more directly provide methods for this if necessary.

as.cl_ultrametric is a generic function which can be used for coercing raw (non-classed) ultrametrics, represented as numeric vectors (of the lower-half entries) or numeric matrices, to ultrametric objects.

Ultrametric objects are implemented as symmetric proximity objects with a dissimilarity interpretation so that self-proximities are zero, and inherit from classes "cl_dissimilarity" and "cl_proximity". See section Details in the documentation for cl_dissimilarity for implications.

Ultrametric objects can also be coerced to classes "dendrogram" and "hclust", and hence in particular use the plot methods for these classes. By default, plotting an ultrametric object uses the plot method for dendrograms.

Value

An object of class "cl_ultrametric" containing the ultrametric distances.

See Also

is.cl_hierarchy

Examples

hc <- hclust(dist(USArrests))
u <- cl_ultrametric(hc)
## Subscripting.
u[1 : 5, 1 : 5]
u[1 : 5, 6 : 7]
## Plotting.
plot(u)

clue documentation built on Sept. 23, 2023, 5:06 p.m.