Simulate elicitation accuracy across question counts — run_accuracy_vs

Runs a robot-agent calibration sweep for each value in q_range and returns the mean absolute error (MAE) between the true and recovered risk parameter. Useful for visualising the accuracy–effort trade-off.

Usage

run_accuracy_vs_questions(
  q_range = 1:16,
  param_values = seq(-1.5, 3.5, by = 1),
  trials_per_param = 10,
  model = "CRRA",
  val_min = 0,
  val_max = 120,
  seed = 42
)

Arguments

q_range: Integer vector of question counts to evaluate (default 1:16).
param_values: Numeric vector of true parameter values to simulate.
trials_per_param: Integer. Repetitions per parameter value.
model: "CRRA" or "CARA".
val_min, val_max: Outcome range.
seed: Random seed for reproducibility.

Value

A data frame with columns n_questions and mae.