Skip to content

Introduction

SymbolicDSGE is a linear DSGE engine with a completely symbolic model specification. Through SymPy, model components are parsed into expressions that can be adjusted, decomposed, and analyzed. This allows things like searching model parameters in a grid, quickly modifying and testing parsed equations, and more; All parsed components of the model support overriding and recompiling. The library is actively being developed, and features are expanding quickly. Current functionality includes:

  • YAML-based model configuration
  • Parser with a SymPy backend
  • linearsolve based solver
  • IRF path/plot generation
  • Simulation
  • Shock generation interface with support for all SciPy distributions
  • Bayesian prior API (Prior, make_prior) with transforms + distribution dispatch
  • Estimation workflows (MLE, MAP, adaptive Metropolis MCMC)
  • Data retrieval helper for FRED API
  • Data transformation functions (HP filters, detrending, etc.)
  • Kalman Filter implementation
  • Monte Carlo pipeline framework with support for user-defined steps, context objects, and result containers
  • OLS, Ridge, Lasso, and Elastic Net regression with coordinate descent solvers and grid search utilities with Monte Carlo regression steps
  • Statistical test utilities with Monte Carlo test steps