Compute Qini Curves for Multi-Arm Treatments — compute_qini_curves_multi

This function computes Qini curves for multi-arm treatment effects using the maq package. It handles various edge cases and provides detailed information about the computation process.

This function computes Qini curves for multi-arm causal forests using doubly robust scores. It handles various edge cases and provides fallback mechanisms to ensure valid outputs.

Usage

compute_qini_curves_multi_arm(
  model,
  tau_hat,
  Y,
  W_multi,
  W.hat = NULL,
  cost = NULL,
  verbose = TRUE
)

compute_qini_curves_multi_arm(
  model,
  tau_hat,
  Y,
  W_multi,
  W.hat = NULL,
  cost = NULL,
  verbose = TRUE
)

Arguments

model: A fitted GRF multi-arm causal forest model.
tau_hat: Matrix of estimated treatment effects (one column per treatment arm comparison).
Y: Vector or matrix of observed outcomes.
W_multi: Factor vector of multi-arm treatment assignments.
W.hat: Optional vector of treatment probabilities (one for each arm).
cost: Optional vector of costs (one for each arm).
verbose: Logical; if TRUE, print diagnostic information during execution.

Value

A list containing:

qini_data: A data frame containing Qini curve data for plotting.
qini_objects: A list of maq objects for each curve, which can be used to compute average gain.

The qini_data has an attribute "imputed" which is TRUE if any curves were imputed with zeros.

A list containing:

qini_data: A data frame containing Qini curve data for plotting.
qini_objects: A list of maq objects for each treatment comparison.

Returns NULL if an error occurs.

Details

The function computes Qini curves for all arms combined, a baseline (no covariates), and each individual treatment arm. It handles cases where some or all Qini objects have zero length or are NULL, extending curves with zeros when necessary.