User Guide

This section contains some useful guides to using MacroStat. For a comprehensive guide to the implemented models, see Model Library, and for the full API, see API reference.

• Getting Started
• Core Model Structure
  • Introduction
  • Module Components
• Differentiability and Jacobians
  • Overview
  • Quick start
  • Direct vs log-space perturbations
  • High-level differentiability check
  • Per-parameter Jacobian comparison
  • Expected disagreements
  • Test model: LINEAR2D
• Parameter constraints example
  • A minimal Parameters subclass
  • What happens when defaults violate the constraint?
  • The happy path: free parameters move, derived parameter follows
  • Autograd consequence: derived parameter has zero gradient
  • Real-world example: GL06INSOUT
  • Closing notes
• Jacobian diagnostics
  • LINEAR2D: direct-space differences
  • GL06SIMEX: log-space on a smooth SFC model
  • GL06INSOUT: localising a step-function disagreement
  • Testing different values of epsilon in the numerical differentiation
  • Decision tree: reading a compare_jacobian_dicts report
• Parameter Estimation and Calibration
  • Overview
  • Quick start
  • Loss functions
  • Calibration best practices
  • Working with EstimationResult
  • Integration with differentiability tools
  • Levenberg-Marquardt optimization
  • Complete calibration examples
  • See also
• Causality Analysis
  • The Causality Module
  • Docstring Causality
• Parameter Space Sampling
  • The Sample Module
• LaTeX Interfaces