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. |