Perform Cost Sensitivity Analysis for QINI Curves

Generates QINI curves across multiple treatment cost scenarios to understand how optimal treatment allocation changes with cost. This allows exploration of policy recommendations under different budget constraints without rerunning the causal forest models.

Usage

margot_qini_cost_sensitivity(
  models,
  costs = c(0.2, 0.5, 1, 2, 5),
  model_names = NULL,
  spend_levels = 0.1,
  baseline_method = "maq_no_covariates",
  verbose = TRUE,
  seed = 12345
)

Arguments

models

List returned by margot_causal_forest(), containing results and optionally full_models.

costs

Numeric vector of treatment costs to evaluate. Default is c(0.2, 0.5, 1, 2, 5) representing a range from cheap to expensive treatments.

model_names

Optional character vector specifying which models to process. Default NULL (all models).

spend_levels

Numeric vector of spend levels for difference gain summaries. Default is 0.1.

baseline_method

Method for generating baseline. See margot_qini() for details.

verbose

Logical; print progress messages (default TRUE).

seed

Integer; base seed for reproducible computations (default 12345).

Value

A list with class "margot_qini_cost_sensitivity" containing:

• results: Named list where each element is a cost scenario containing QINI results from margot_qini()

• costs: The cost values used

• models_processed: Names of models that were processed

• summary: Data frame summarizing key metrics across costs

Details

This function systematically varies the treatment_cost parameter in margot_qini() to show how QINI curves and optimal treatment fractions change with cost.

Lower costs (e.g., 0.2) represent scenarios where treatment is cheap relative to budget, resulting in steeper QINI curves and larger optimal treatment fractions. Higher costs (e.g., 5) represent expensive treatments where only the highest-effect individuals justify treatment.

The summary table includes the optimal treatment fraction and expected gain at key spend levels for each cost scenario, facilitating comparison across different budget constraints.

Examples

if (FALSE) { # \dontrun{
# Basic cost sensitivity analysis
cost_sens <- margot_qini_cost_sensitivity(
  causal_forest_results,
  costs = c(0.5, 1, 2)
)

# View summary of optimal treatment fractions
print(cost_sens$summary)

# Plot results for a specific model across costs
margot_plot_qini_cost_sensitivity(cost_sens, model_name = "model_anxiety")

# Detailed analysis for specific models
cost_sens_detailed <- margot_qini_cost_sensitivity(
  causal_forest_results,
  costs = seq(0.2, 3, by = 0.2), # Fine-grained cost grid
  model_names = c("model_anxiety", "model_depression"),
  spend_levels = c(0.05, 0.1, 0.2, 0.4) # More spend levels
)
} # }