Plot counterfactual outcome densities — plot_counterfactual

Visualizes the distribution of the outcome under different treatment regimes. Two modes: (1) shift-and-predict using trained engression models (default), which shows marginal outcome distributions P(Y|do(A=a)); (2) weighted densities using the learned Riesz representers (use_weights = TRUE), which shows all mediation counterfactual regimes including cross-world quantities like Y(1, M(0)).

Usage

plot_counterfactual_density(
  result,
  n_samples = 1,
  regimes = NULL,
  use_weights = FALSE
)

Arguments

result: An object of class dma_result.
n_samples: Number of samples to draw per observation (for shift-and-predict).
regimes: A named list of shift functions or numeric values for treatment. Default uses the d0/d1 from the original call. Only used when use_weights = FALSE.
use_weights: If TRUE, uses the learned density ratios (Riesz representers) to estimate counterfactual distributions via importance weighting. This shows all mediation-relevant regimes, not just the two marginal treatment regimes. Default is FALSE.

Value

A ggplot2 object.