Skip to content

Frequently Asked Questions

Common questions about the NHRA Game Theory toolkit.

General

What is this project?

The NHRA Game Theory toolkit is a stylised mechanism model for analysing healthcare funding negotiations under the National Health Reform Agreement (NHRA). It uses game theory to model strategic interactions between Commonwealth and State governments.

Not a Forecast

This is a policy reasoning tool, not a predictive model. It helps understand incentive structures and compare policy directions.

Who is this for?

  • Health policy researchers
  • Healthcare economists
  • Policy analysts
  • Academic researchers studying mechanism design

What programming language is used?

Python 3.10+ with NumPy, Pandas, and optional JAX acceleration.

Technical Questions

How do I install the toolkit?

# Using Poetry (recommended)
git clone https://github.com/edithatogo/nhra_game.git
cd nhra_game
poetry install
poetry shell

See the Usage Guide for more details.

What are the system requirements?

  • Python 3.10, 3.11, 3.12, or 3.13
  • ~500MB disk space
  • 4GB RAM recommended for Monte Carlo simulations

How do I run a simulation?

from nhra_gt.engine import Params, run_simulation

results = run_simulation(years=10, n_samples=100, params=Params())

What does "strict mypy mode" mean?

The codebase uses mypy in strict mode, requiring:

  • Type hints on all functions
  • No implicit Any types
  • Explicit return types

This ensures type safety and catches bugs early.

Game Theory Questions

What is a Nash equilibrium?

A Nash equilibrium is a strategy profile where no player can improve their payoff by unilaterally changing their strategy. It represents a stable outcome where each player's strategy is a best response to others.

Why use game theory for healthcare policy?

Healthcare funding involves strategic interactions between multiple parties (Commonwealth, States, hospitals) with different objectives. Game theory provides a rigorous framework to analyse these interactions and predict equilibrium outcomes.

How are payoff matrices constructed?

Payoff matrices are parameterised by state variables (pressure, efficiency gap, etc.). The formulas are stylised representations based on:

  • Economic incentives (funding share, cost shifting)
  • Political factors (salience, electoral cycles)
  • Operational constraints (discharge delays, capacity)

See Game Theory Models for detailed payoff formulas.

What is a "stage game"?

Each year of the simulation, players engage in multiple stage games representing different decision points:

  • Definition (what counts as efficient?)
  • Bargaining (converge or defer?)
  • Cost shifting (invest or shift?)
  • etc.

The outcomes of these games affect state variables that carry forward to the next year.

Validation Questions

How is the model validated?

The validation suite includes:

  1. Sensitivity Analysis: Sobol global sensitivity to identify key parameters
  2. Recursive Backtesting: Rolling horizon validation against historical patterns
  3. Mechanism Validation: Verify game-theoretic properties hold
  4. Unit Tests: Comprehensive pytest suite with property-based testing

See Validation Documentation for details.

What are the model's limitations?

Key limitations include:

  • Stylised payoffs (not empirically estimated)
  • Two-player simplification of multi-stakeholder dynamics
  • No geographic heterogeneity
  • Static games (no repeated game dynamics)

See Requirements & Scope for full discussion.

How do I report a bug?

Open an issue on GitHub with:

  • Description of the issue
  • Steps to reproduce
  • Expected vs actual behaviour
  • Python version and OS

Data & Parameters

Where do the parameters come from?

All parameters must be grounded in publicly accessible sources. The parameter registry (context/04_parameter_registry.csv) tracks:

  • Value and units
  • Source type (primary, secondary, calibrated, assumed)
  • Public URL for traceability
  • Plausible range for sensitivity analysis

Can I use my own parameters?

Yes! Create a custom Params object:

from nhra_gt.engine import Params

custom = Params(
    cost_shifting_intensity=0.3,
    political_salience=0.6,
    # ... other parameters
)

How do I add new data sources?

  1. Add the parameter to context/04_parameter_registry.csv
  2. Include the public URL and source details
  3. Run python scripts/check_parameters_grounded.py to validate

Contributing

How can I contribute?

See the Contributing Guide for:

  • Development setup
  • Code style requirements
  • Pull request process

Is the code open source?

Yes, licensed under Apache 2.0. See LICENSE.

Still have questions?

Open an issue on GitHub or check the full documentation.