This directory contains example scripts demonstrating how to use heron with various tools and frameworks.

File: bilby_integration_example.py

This example shows how to use heron's Gaussian process-based waveform models with the bilby gravitational wave inference library. The key features are:

• Using HeronWaveformGenerator to create bilby-compatible waveforms from heron models
• Using HeronGravitationalWaveTransient likelihood that properly handles waveform uncertainty
• Incorporating model uncertainty from Gaussian process regression into parameter estimation

The example requires both heron and bilby to be installed:

Then you can run the example:

Note: You'll need to uncomment the main() or compare_with_without_uncertainty() function call at the bottom of the script to actually run an analysis.

The example demonstrates two main components:

1. HeronWaveformGenerator: A bilby-compatible waveform generator that wraps heron waveform models and propagates uncertainty information.

2. HeronGravitationalWaveTransient: A bilby likelihood that incorporates waveform model uncertainty into the likelihood calculation by adding the model covariance to the detector noise covariance.

When include_model_uncertainty=True (the default), the likelihood accounts for uncertainty in the waveform predictions that comes from the Gaussian process regression. This leads to more conservative parameter estimates that properly reflect the model's confidence in different regions of parameter space.

The uncertainty is propagated through:

• Antenna response projection (F+ and Fx)
• FFT operations for frequency domain analysis
• Matched filtering calculations

The compare_with_without_uncertainty() function shows how to compare results with and without model uncertainty, demonstrating the impact of including this information in the analysis.