Skip to content

Bootstrap Configuration (Auto-Convergence)

The bootstrap_* structs (e.g., bootstrap_thresholds) support the following nested parameters to control the convergence algorithm:

Field Type Default (Thr/CI/Rank) Description
B_start int 100 / 100 / 50 Initial number of bootstrap iterations.
B_step int 100 / 200 / 25 Iterations to add in each step.
B_end int 10000 / 20000 / 2500 Maximum number of iterations.
n_trials int 25 / 30 / 15 Number of independent trials per step to check stability.
convergence_tolerance double 0.01 / 0.03 / 0.005 Max allowed variation (e.g., 0.005 = 0.5%).
smoothing_window int 3 / 3 / 3 Window size for moving average smoothing.
convergence_streak_needed int 3 / 3 / 3 Consecutive steps required to pass tolerance.
min_steps_for_convergence_check int 1 Minimum steps before checking convergence.

👉 Convergence Modes and Troubleshooting