SelectnrClusters: Select the number of clusters via silhouette widths

In MosaiClusteR: An Umbrella Framework for Multi-Source and Multi-Omics Clustering

Select the number of clusters via silhouette widths

Description

For each element of List, partitions around medoids (cluster::pam) are computed over a range of cluster numbers and the average silhouette width is recorded. The number of clusters maximising the average silhouette width is reported per source and for the across-source average. The analytical result (the silhouette table and the chosen number of clusters) is returned and is testable without producing any plot.

Usage

SelectnrClusters(
  List,
  type = c("data", "dist", "pam"),
  distmeasure = c("tanimoto", "tanimoto"),
  normalize = c(FALSE, FALSE),
  method = c(NULL, NULL),
  nrclusters = seq(5, 25, 1),
  names = NULL,
  StopRange = FALSE,
  plottype = "none",
  location = NULL
)

Arguments

List	A list of data matrices (objects in rows), distance matrices, or cluster::pam outputs, depending on type.
type	One of "data" (compute a distance first via Distance), "dist" (elements are dissimilarity matrices) or "pam" (elements are pam objects).
distmeasure	Distance measure(s) used when type = "data".
normalize, method	Normalisation controls passed to Distance.
nrclusters	Sequence of cluster numbers to evaluate.
names	Optional names for the sources; used in output labels.
StopRange	Logical; if FALSE and dissimilarities fall outside [0, 1] they are range-normalised for comparability.
plottype	One of "none" (default; no plot, safe for headless runs), "new", "pdf" or "sweave". Any value other than "none" attempts to draw the silhouette curve; drawing is guarded so that a missing graphics device does not abort the computation.
location	File location (without extension) used when plottype = "pdf". Default NULL.

Value

A list with two elements: Silhoutte_Widths (a matrix of average silhouette widths per source and their mean) and Optimal_Nr_of_CLusters (a one-row data frame with the silhouette optimal number of clusters per source and overall).

References

Kaufman, L. and Rousseeuw, P. J. (1990). Finding Groups in Data: An Introduction to Cluster Analysis. Wiley, New York.

Examples

## Not run: 
data(mosaic_toy)
sel <- SelectnrClusters(mosaic_toy$List, type = "data",
                        distmeasure = c("euclidean", "euclidean"),
                        nrclusters = seq(2, 5), plottype = "none")
sel$Optimal_Nr_of_CLusters

## End(Not run)

MosaiClusteR documentation built on Oct. 25, 2026, 1:08 a.m.