1. Understanding a rise_model object

Every modern factory – dsge_model(...), rfvar_model(...), svar_model(...), proxy_svar_model(...), prfvar_model(...), dsge_var_model(...) – returns an object of the same class: rise_model. The fields below are the canonical ones; the engines (solve, filter, estimate, forecast, irf, decompositions) read these and only these, regardless of which factory produced the object.

1.1. Ordering and information about atoms

1.1.1. Endogenous variables

For DSGE-shaped models, endogenous variables are ordered by group, and within each group alphabetically:

static variables – variables dated in period t only.
purely predetermined variables – variables dated in periods t and t-1.
predetermined and forward-looking variables – variables dated in periods t, t-1 and t+1.
purely forward-looking variables – variables dated in periods t and t+1.

For VAR-shaped models (reduced-form, structural, proxy SVAR, panel, DSGE-VAR) the endogenous variables are in the order they were passed to the factory. Panel models additionally append a per-unit suffix during expansion (GROWTH_US, GROWTH_CA, …).

1.1.2. Exogenous variables

Alphabetical order.

1.1.3. Parameters

Alphabetical order.

1.1.4. Observables

Alphabetical order.

1.2. Information on model equations

get(m, 'equations') returns the equations as they were written, together with their labels (the quoted strings that appear before each equation in the model file). The equations are the parser-canonical form; the original raw text of the model file is not preserved.

VAR-family models do not have model-file equations in the same sense; get(m, 'equations') returns the canonical structural / reduced-form representation the parser built from the factory arguments.

1.3. Provenance: which RISE version built the model

Every constructed model records the RISE version that produced it, and – once estimated – the version that estimated it. Query it through the get interface; the keys also appear in the no-argument get(m) catalog of gettable elements:

m = dsge_model('fs2000');

get(m, 'rise_version')            % struct: .construction / .estimation
% ans.construction -> '20260626'  (version that built m)
% ans.estimation   -> ''          (empty: not estimated yet)

get(m, 'construction_version')    % '20260626'
get(m, 'estimation_version')      % ''

The same values are reachable programmatically as m.construction_version, m.estimation_version and m.rise_version. These three are kept out of the model’s default display to avoid clutter, but the display does carry a compact version triple ordered [installed | construction | estimation] – the currently installed toolbox alongside the two stamps – so a mismatch (a model built or estimated under a different RISE than the one now on the path) is visible at a glance.

For options specifically, get(m, 'options') returns this model’s current settable options and their values – the get-side companion to set. The full catalog of option names with their default values is set(rise.engine.models.rise_model.empty). The gettable catalog is the no-argument get(m).

The construction stamp is set when the object is built and the estimation stamp when estimate runs, so a model built under one version and estimated under a later one reports both faithfully:

mest = estimate(m, 'data', db, 'estim_priors', priors, ...);
get(mest, 'estimation_version')   % version that ran the estimation

Both stamps are frozen into model_data and therefore survive a save/load round-trip – loading a model under a newer RISE does not overwrite the version that originally built it. summary(m) prints them as Built with RISE / Estimated with RISE lines.

This is distinct from rise.version, which reports the currently installed toolbox rather than any particular model. Models saved before provenance stamping was introduced load unchanged and report '' (unknown) for the missing stamps.

1.4. Information on model results

1.4.1. First-order structural form

The matrices of the parsed first-order structural form are reachable via extract_first_order_structure. They include the contemporaneous, lead and lag Jacobians used to feed the perturbation solver (DSGE-shaped models) or the structural-form matrices A0, A1, …, Ap (SVAR-shaped models).

1.4.2. Solution results

After solve succeeds, the solution lives in m.state_space{1} (one entry per parameterization) and includes the regime-specific transition matrices, the shock-impact matrices, and the steady state. Use print_solution(m) for a human-readable view:

print_solution(m)
print_solution(m, {'pi','y'})        % restrict to a subset

The shape of state_space is the same across factories – the unified state-space representation described in Modern architecture.

1.4.3. Posterior maximisation results

After estimate succeeds:

m_est.estimation.posterior_maximization.mode
m_est.estimation.posterior_maximization.mode_stdev
m_est.estimation.posterior_maximization.hessian
m_est.estimation.posterior_maximization.vcov
m_est.estimation.posterior_maximization.log_post
m_est.estimation.posterior_maximization.log_lik
m_est.estimation.posterior_maximization.log_prior
m_est.estimation.posterior_maximization.log_marginal_data_density_laplace

The field names and shapes are the same across factories.

See Estimation for the full surface.