Tutorials

Use these tutorials as a guided path through the core workflows.

• Quickstart (Bandit)
• Quickstart (MDP)
• Diagnostics
• Confidence Intervals
• Sensitivity Analysis
• Long-Horizon MIS vs IS
• Proximal OPE (Advanced)

Notebook gallery

The repository notebooks are paired with .py scripts via Jupytext, so you can review diffs cleanly and run them as scripts. They include end-to-end pipelines, diagnostics deep dives, and advanced identification tutorials.

Suggested entry points:

• 00_introduction.ipynb — estimand-first tour with report export
• 01_estimands_and_assumptions.ipynb — assumptions enforcement and overlap stress test
• 02_bandit_ope_walkthrough.ipynb — diagnostics-driven bandit OPE
• 03_mdp_ope_walkthrough.ipynb — trajectory estimators and horizon effects
• 04_confidence_intervals_and_hcope.ipynb — CIs and high-confidence bounds
• 05_sensitivity_analysis_bandits.ipynb — bandit sensitivity curves
• 06_sensitivity_unobserved_confounding.ipynb — sequential Gamma bounds
• 07_proximal_ope_confounded_pomdp.ipynb — proximal OPE with proxies
• 08_estimator_selection_and_debugging.ipynb — selection heuristics and playbook
• 10_bandit_ope_end_to_end.ipynb — research-grade workflow with sensitivity
• 11_mdp_long_horizon_comparison.ipynb — long-horizon estimator comparison