Significant Testing of Clustering Results
Uses the SigClust K-Means algorithm to assess significance of clustering results.
sigclust(x, k, nsim, nrep = 1, labflag = 0, label = 0, icovest = 2)
x |
data matrix, samples are rows and features are columns |
k |
cluster size to test against |
nsim |
number of simulations |
nrep |
See |
labflag |
See |
label |
true class label. See |
icovest |
type of covariance matrix estimation |
This function is a wrapper for the original sigclust::sigclust(), except
that an additional parameter k is allows testing against any number of
clusters. In addition, the default type of covariance estimation is also
different.
An object of class sigclust. See sigclust::sigclust() for
details.
Hanwen Huang: hanwenh@email.unc.edu; Yufeng Liu: yfliu@email.unc.edu; J. S. Marron: marron@email.unc.edu
Liu, Yufeng, Hayes, David Neil, Nobel, Andrew and Marron, J. S, 2008, Statistical Significance of Clustering for High-Dimension, Low-Sample Size Data, Journal of the American Statistical Association 103(483) 1281–1293.
data(hgsc) dat <- hgsc[1:100, 1:50] nk <- 4 cc <- consensus_cluster(dat, nk = nk, reps = 5, algorithms = "pam", progress = FALSE) cl.mat <- consensus_combine(cc, element = "class") lab <- cl.mat$`4`[, 1] set.seed(1) str(sigclust(x = dat, k = nk, nsim = 50, labflag = 1, label = lab))
Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.