Expand description
§solow-statespace
Linear-Gaussian state-space models and SARIMAX estimation.
The crate provides two layers:
kalman— a time-invariantStateSpacewith an exact Kalmanfilter(returning the Gaussian log-likelihood and the filtered/predicted state sequences) and a fixed-intervalsmoother.sarimax— aSarimaxseasonal ARIMA estimator built on the Kalman filter and fit by maximum likelihood. The AR/MA and seasonal polynomials are mapped into the Harvey companion-form state space, stationarity/invertibility are enforced by the Monahan reparametrization, and the model is optimized with BFGS over-loglike.
use ndarray::Array1;
use solow_statespace::{Sarimax, SarimaxOrder};
// A short AR(1)-like series.
let y = Array1::from_vec(vec![
0.2, 0.5, 0.1, -0.3, 0.4, 0.8, 0.3, -0.1, 0.0, 0.6, 0.9, 0.2, -0.4, 0.1,
]);
let model = Sarimax::new(y, SarimaxOrder::new(1, 0, 0)).unwrap();
let res = model.fit().unwrap();
assert_eq!(res.params.len(), 2); // [ar.L1, sigma2]Re-exports§
pub use dynamic_factor::DynamicFactor;pub use dynamic_factor::DynamicFactorResults;pub use kalman::FilterOutput;pub use kalman::StateSpace;pub use mvkalman::MvStateSpace;pub use sarimax::Sarimax;pub use sarimax::SarimaxOrder;pub use sarimax::SarimaxResults;pub use structural::Level;pub use structural::UcResults;pub use structural::UcSpec;pub use structural::UnobservedComponents;
Modules§
- dynamic_
factor - Dynamic-factor model with a single common factor, fit by maximum likelihood.
- kalman
- Linear-Gaussian Kalman filter and smoother.
- mvkalman
- Multivariate linear-Gaussian Kalman filter.
- sarimax
- SARIMAX(p, d, q)(P, D, Q, s) estimation by maximum likelihood.
- structural
- Unobserved-components (structural time-series) models, fit by maximum likelihood.