Choosing a Sampler¶
Coming Soon
This guide is being written. Sampler Python bindings are also in development.
Quick Reference¶
| Use Case | Sampler |
|---|---|
| Uncertainty quantification | Metropolis-Hastings |
| Model comparison | Dynamic Nested Sampling |
| Just posterior samples | Metropolis-Hastings |
| Need evidence/Bayes factors | Dynamic Nested Sampling |
When to Sample vs Optimize¶
Use optimisers when: - Point estimate is sufficient - Speed is critical - Don't need uncertainty
Use samplers when: - Need confidence intervals - Comparing models - Want full posterior