Visualise a shape-sensitivity sweep

Produce one of four diagnostic plots from the raw data frame returned by janusplot_shape_sensitivity():

"confusion_fine"

24 x (|shapes|) confusion matrix at the fine category level — rows = ground truth, columns = predicted, cells coloured by P(pred | truth).

"confusion_archetype"

7 x 7 confusion matrix at the archetype level.

"accuracy_grid"

per-shape heatmap of archetype-level accuracy across the (n, sigma) design.

"recovery_curves"

accuracy as a function of sigma, one line per sample size, faceted by shape.

Usage

janusplot_shape_sensitivity_plot(
  results,
  type = c("confusion_fine", "confusion_archetype", "accuracy_grid", "recovery_curves")
)

Arguments

results

Data frame from janusplot_shape_sensitivity() or the precomputed shape_sensitivity_demo.

type

One of "confusion_fine", "confusion_archetype", "accuracy_grid", or "recovery_curves".

Value

A ggplot2::ggplot object.

See also

Other shape-sensitivity: janusplot_shape_sensitivity(), janusplot_shape_sensitivity_shapes(), janusplot_shape_sensitivity_summary()

Examples

data("shape_sensitivity_demo", package = "janusplot")
janusplot_shape_sensitivity_plot(shape_sensitivity_demo,
                                 "recovery_curves")