ClusterStability: Compute cluster stability of MCAP output

In btaschler/mcap: Model-Based Clustering in High Dimensions via Adaptive Projections

Description

Compute cluster stability w.r.t. pairs of data points over B random subsamples of size m < n (where n is the total sample size). Default for m=0.75n. Clustering is done using a full covariance Gaussian mixture model. Measure of cluster stability is based on the adjusted Rand index of shared points given two subsets of the data.

Usage

ClusterStability(xx, k, B = 10, frac_subsample = 0.75)

Arguments

xx	The data matrix (n x p).
k	The number of clusters.
B	The number of subsampled data sets to be used (default: B=10). Note that if B is too small (<10), the variance of the stability estimate increases.
frac_subsample	Fraction of total samples to be used in each subsample (default: 0.75).

Value

stab	Measure of cluster stability.

Author(s)

Bernd Taschler: [e-mail]

Sach Mukherjee: [e-mail]

See Also

MCAPfit, OptDimClusterStability, GMMwrapper, adjustedRandIndex

Examples

  ## standard Normal matrix, 2 clusters:
  ClusterStability(xx = matrix(rnorm(1e3), 100), k = 2)
  
  ## Not run: 
  ## large input matrix 
  ClusterStability(xx = matrix(rnorm(1e3*1e4), 1e3), k = 5, B = 20)
  
## End(Not run)

btaschler/mcap documentation built on May 26, 2019, 1:31 a.m.