Solve and estimate Dynamic Stochastic General Equilibrium models (including the New York Fed DSGE)
nelder_mead optimizerAltPolicy) and alternative
scenarios (AbstractScenario)plot_parameters, plot_history_and_forecast,
plot_forecast_comparison, hair_plot, plot_shock_decomposition,
plot_impulse_response, plot_altpolicies, and plot_scenario
get_data_filename
Setting with key data_id
cond_id from Setting{String} to Setting{Int}
inpath(m, "data") to inpath(m, "raw")
:marginal_L (marginal likelihood) field to Kalman typeMM and VVall fields from Measurement type:states, :shocks, and :stdshocks
PseudoObservableMapping type (and field in System type) to
PseudoMeasurement
m.pseudo_observables and m.pseudo_observable_mappings fields to
AbstractModel subtypesNullable. Instead, if no
pseudo-measurement equation is implemented, the fields in the model object
are empty dictionariesmeans_bands_all to compute_meansbands
meansbands_matrix_all to meansbands_to_matrix
This release of the FRBNY DSGE.jl package implements Sequential Monte Carlo (SMC) sampling as an alternative to Metropolis Hastings Markov Chain Monte Carlo sampling. The SMC algorithm implemented here is based upon Edward Herbst and Frank Schorfheide's paper "Sequential Monte Carlo Sampling for DSGE Models" and the code accompanying their book Bayesian Estimation of DSGE Models. More information and the original MATLAB scripts that this code replicates can be found at Frank Schorfheide's website. Currently, FRBNY's implementation of SMC works on the small-scale New Keynesian DSGE model presented in Sungbae An and Frank Schorfheide's paper "Bayesian Analysis of DSGE Models". FRBNY is currently working on extending the code so that SMC may be used with medium-scale DSGE models. This and other extensions of the DSGE model code may be released in the future at the discretion of FRBNY. Comments and suggestions are welcome, and best submitted as either an issue or a pull request to this branch.
MultivariateOptimizationResults type requires f_increased fieldMersenneTwister must be constructed with a seed:simulated_annealing, :LBFGS, and
:combined_optimizer (which alternates between simulated annealing and
LBFGS).PseudoObservable type and the pseudo_measurement function, which
defines pseudo-observables (linear combinations of states which are not
observed) for each model, e.g. the output gap.forecast_one and means_bands_all; see the
forecasting and
means and bands
for more details.Observable type; replaced the data_series and data_transforms
fields in the model type definitions with
observable_mappings::OrderedDict{Symbol, Observable}, which is initialized
in init_observable_mappings!.kalman_filter has been broken out into
StateSpaceRoutines.jl.estimate now saves only parameter draws, not the associated state-space
matrices or the last filtered states for each draw.betabar to use
m[:σ_c] instead of σ_ω_star
betabar
transpose for Parameters so that matrix division (i.e. the
(\) operator) no longer throws a warning