Skip to contents

Plot Best-Subset Optimization Diagnostic from a mosaic_subset_optimization Object

Source: R/plot_model_subset_optimization.R
plot_model_subset_optimization.Rd

Renders a four-panel diagnostic from a mosaic_subset_optimization object produced by optimize_ensemble_subset. Each panel plots a diagnostic against the candidate subset size N (top-N members by likelihood): (1) the objective score, (2) \(R^2\) for cases and deaths, (3) the bias ratio for cases and deaths (with a dotted reference line at bias = 1), and (4) the effective sample size (ESS). The preliminary (tier/diagnostics) subset size is marked with a dashed grey vertical line and the optimizer-selected size with a solid black vertical line plus a point on the score panel.

Usage

plot_model_subset_optimization(
  subset_opt,
  output_dir,
  file_prefix = "subset_optimization",
  title_label = NULL,
  verbose = TRUE
)

Arguments

subset_opt

A mosaic_subset_optimization object returned by optimize_ensemble_subset. Its $evaluation_table carries the per-N diagnostics; the scalar fields $optimal_n, $diagnostics_n, $objective, and $central_method drive the markers and labels. The object is consumed directly; the diagnostics are not re-derived from the on-disk CSV.

output_dir

Character. Directory where plots are saved. Created if it does not exist.

file_prefix

Character. Prefix used in output filenames: <file_prefix>_diagnostic.pdf and <file_prefix>_diagnostic.png. Default "subset_optimization".

title_label

Character or NULL. Leading label used in the plot title. When NULL (default) a title derived from the objective and central method is used.

verbose

Logical. Print progress messages. Default TRUE.

Value

Invisibly returns the ggplot object.

Details

The PDF is rendered with grDevices::cairo_pdf when the cairo device is available (capabilities("cairo")), so the proper glyphs \(R^2\), \(\rightarrow\), and \(\Delta\) are encoded correctly. When cairo is unavailable the function falls back to the default PDF device with ASCII-safe labels (R2, ->, delta). The PNG is always rendered with the standard raster device.

Semantic colors for the cases and deaths series come from mosaic_colors and the panel theming from theme_mosaic, both with graceful fallbacks if unavailable.

See also

optimize_ensemble_subset to compute the optimization; plot_model_ensemble for the ensemble prediction plots.

Examples

if (FALSE) { # \dontrun{
subset_opt <- optimize_ensemble_subset(ensemble, likelihoods)
plot_model_subset_optimization(subset_opt, output_dir = "diagnostics")
} # }