Plot pairwise distance benchmark comparisons

Draws an annotated triangular heatmap of the pairwise diagnostics from [benchmark_mdist()].

Usage

# S3 method for class 'MDistBenchmark'
autoplot(
  object,
  metric = c("relative_distance", "mad", "alienation", "mds_congruence", "ari"),
  cluster_method = NULL,
  digits = 2,
  ...
)

Arguments

object

An object returned by [benchmark_mdist()].

metric

Character string selecting `"mad"`, `"relative_distance"`, `"mds_congruence"`, `"alienation"`, or `"ari"`. A specific ARI column, such as `"ari_pam"`, can also be supplied.

cluster_method

Optional clustering method used when `metric = "ari"`. If `NULL`, all available ARI metrics are shown, with one facet per clustering method.

digits

Number of decimal places used for cell labels.

...

Currently unused.

Value

A `ggplot` object.

Details

The default fill gradient runs from coral (`#E76F51`) for lower values to blue (`#008CFF`) for higher values, with regular-weight white cell labels. To use another palette, add a fill scale to the returned plot. Self-comparisons on the main diagonal are left blank. ARI plots use fixed colour limits of 0 and 1, so the same ARI has the same colour across plots. Negative ARIs retain their numerical labels but use the low-end colour.