Bootstrap Confidence-Interval Engine Forensics¶
- Task ID:
computer_science.bci_forensics - Domain:
computer_science - Subdomain:
statistical_computing - Status:
test - Tags:
bootstrap,confidence_intervals,statistical_computing,uncertainty_quantification,simulation_diagnostics,forensic_audit
Public Summary¶
This page is generated from task metadata and selected public-safe excerpts.
Example B1 Prompt Excerpt¶
You are given an anonymous bootstrap confidence-interval engine forensics instance. Build a reproducible analysis from the files in `data/`. Write the reproducible script as `analysis.py` at the workspace root, and write every artifact under `results/`.
This task is not asking you to invent a new interval from scratch. It gives candidate interval engines and asks you to audit them. Use the supplied observed samples, calibration replications, resampling panels, and anonymous engine endpoint arrays exactly as provided.
## Scientific Background
Bootstrap confidence intervals can look plausible while failing in specific regimes. Common failures include:
- a simple quantile endpoint rule that works near symmetric smooth statistics but loses balance in skewed or nonlinear statistics;
- a reflected endpoint rule that can overreact to skew and produce poor tail allocation;
- a discrete or bounded statistic where endpoint conventions and ties matter;
- a corrected endpoint rule whose shape correction has the wrong direction;
Notes¶
- This page is a generated site artifact.
- Higher-level prompt details and internal benchmark specifics may remain intentionally undisclosed.