channel_perm: Clustering statistics with channel-wise permutations

In MortenKrebs/diagtraject: Mining of Diagnosis Trajectories

Description

Calculates clustering statistcs for the clustering of the (weighted) sum of a list of dissimilarity matrices. Performs (stratified) permutations and calculates clustering statistics for permuted datasets.

Usage

channel_perm(
  dist_list,
  B,
  maxclust = NULL,
  nclust = NULL,
  strat.var = NULL,
  stat = c("ASW", "CH"),
  c_weight = rep(1, length(dist_list))
)

Arguments

dist_list	list of dissimilarities in either format 'dist' or as a symmetric n x n matrix.
B	number of permutations
maxclust	maximum number of clusters
nclust	number of clusters. Ignored if maxclust is specified.
strat.var	optional factor of length n by which the permutation is stratified.
stat	vector of strings with cluster statistics to be calculated. Possible stats: "ASW" Average silhouette width. "CH" Calinsky Harabasz index on squared distances.
c_weight	Optional vector of numerics with channel weights.

Details

...

Value

Object of class "ch_perm.stats".

Examples

p <- channel_perm(d_imp_UD2,B = 50, maxclust = 15,strat.var = as.vector(TraMineR::seqlength(seq_mis[[1]])),c_weight = weigth, stat="ASW")

MortenKrebs/diagtraject documentation built on March 6, 2021, 10:54 a.m.