Assess Clustering Stability Using Bootstrap Resampling

Evaluates the stability of MONECA clustering results for a given small.cell.reduction parameter using bootstrap resampling. More stable parameters produce consistent clustering across data subsamples.

Usage

assess_clustering_stability(
  mx,
  small.cell.reduction,
  cut.off = 1,
  n_bootstrap = 50,
  sample_fraction = 0.8,
  segment.levels = 3
)

Arguments

mx

Mobility matrix for analysis.

small.cell.reduction

Small cell reduction parameter to test.

cut.off

Edge weight threshold for network construction. Default is 1.

n_bootstrap

Number of bootstrap samples to generate. Default is 50.

sample_fraction

Fraction of data to use in each bootstrap sample. Default is 0.8.

segment.levels

Number of hierarchical levels for clustering. Default is 3.

Value

Numeric stability score between 0 and 1, where higher values indicate more stable clustering results.

Details

The function performs bootstrap resampling by: 1. Creating multiple subsamples of the mobility matrix 2. Running MONECA clustering on each subsample 3. Extracting clustering signatures (segment memberships) 4. Computing stability metrics based on clustering consistency

The stability score combines multiple measures including: - Consistency of segment assignments across bootstrap samples - Stability of hierarchical structure - Robustness of clique detection

See also

auto_tune_small_cell_reduction, moneca

Examples

if (FALSE) { # \dontrun{
# Generate sample data
mobility_data <- generate_mobility_data(n_classes = 6, seed = 123)

# Assess stability for different parameters (uses default n_bootstrap = 50)
stability_0 <- assess_clustering_stability(mobility_data,
                                          small.cell.reduction = 0)
stability_5 <- assess_clustering_stability(mobility_data,
                                          small.cell.reduction = 5)

cat("Stability with parameter 0:", stability_0, "\n")
cat("Stability with parameter 5:", stability_5, "\n")

# For more thorough assessment, increase n_bootstrap
stability_thorough <- assess_clustering_stability(mobility_data,
                                                  small.cell.reduction = 2,
                                                  n_bootstrap = 100)
} # }