element_agreement {ClustAssess}

Element-Wise Average Agreement Between a Set of Clusterings

Description

Inspect how consistently of a set of clusterings agree with a reference clustering by calculating their element-wise average agreement.

Usage

element_agreement(
  reference_clustering,
  clustering_list,
  alpha = 0.9,
  r = 1,
  rescale_path_type = "max",
  ppr_implementation = "prpack",
  dist_rescaled = FALSE,
  row_normalize = TRUE,
  ncores = 1
)

Arguments

reference_clustering

The reference clustering, that each clustering in clustering_list is compared to. It can be either:

• A numeric/character/factor vector of cluster labels for each element.

• A samples x clusters matrix/Matrix::Matrix of nonzero membership values.

• An hclust object.

clustering_list

The list of clustering results, each of which is either:

• A numeric/character/factor vector of cluster labels for each element.

• A samples x clusters matrix/Matrix::Matrix of nonzero membership values.

• An hclust object.

alpha

A numeric giving the personalized PageRank damping factor; 1 - alpha is the restart probability for the PPR random walk.

r

A numeric hierarchical scaling parameter.

rescale_path_type

A string; rescale the hierarchical height by:

• "max" : the maximum path from the root.

• "min" : the minimum path form the root.

• "linkage" : use the linkage distances in the clustering.

ppr_implementation

Choose a implementation for personalized page-rank calculation:

• "prpack": use PPR algorithms in igraph.

• "power_iteration": use power_iteration method.

dist_rescaled

A logical: if TRUE, the linkage distances are linearly rescaled to be in-between 0 and 1.

row_normalize

Whether to normalize all rows in clustering_result so they sum to one before calculating ECS. It is recommended to set this to TRUE, which will lead to slightly different ECS values compared to clusim.

ncores

the number of parallel R instances that will run the code. If the value is set to 1, the code will be run sequentially.

Value

A vector containing the element-wise average agreement.

References

Gates, A. J., Wood, I. B., Hetrick, W. P., & Ahn, Y. Y. (2019). Element-centric clustering comparison unifies overlaps and hierarchy. Scientific reports, 9(1), 1-13. https://doi.org/10.1038/s41598-019-44892-y

Examples

reference.clustering = iris$Species
clustering.list = list()
for (i in 1:20){
  clustering.list[[i]] = kmeans(iris[,1:4], 3)$cluster
}
element_agreement(reference.clustering, clustering.list)