Heatmap of a sensitivity metric across the scenario grid — plot_sensitivity

Draws a coverage x confounding-strength heatmap of a chosen metric for one estimator, at a fixed number of confounders. Useful for reading off where a method's bias/RMSE blows up.

Usage

plot_sensitivity_heatmap(
  sens,
  metric = "rmse",
  method = "IV2SLS",
  k = NULL,
  title = NULL
)

Arguments

sens: Object returned by gan_sensitivity().
metric: Column of sens$summary to map (e.g. "rmse", "abs_bias", "power"). Default "rmse".
method: Estimator to display. Default "IV2SLS".
k: Number of confounders to slice on. Default: first in the grid.
title: Optional plot title.

Value

Invisibly returns NULL; called for the side effect of drawing a sensitivity heatmap.

Examples

sens <- gan_sensitivity(NULL, conf_grid = c(0.5, 0.8),
  coverage_grid = c(0.3, 0.7), n_iter = 2, n_samples = 100,
  n_features = 5)
plot_sensitivity_heatmap(sens)