Dynare Output Variables
This page is under construction. It is meant at documenting existing output variables and at discussing a better organization of these.
At the end of a successful run, Dynare stores the following variables in the file <filename>_results.mat:
M_: Matlab structure containing internal variables used to keep track of variables, parameters and other characteristics of the model
options_: Matlab structure containing the options used in the model run (default options or user specified options)
oo_: Matlab structure containing various results obtained by Dynare
To be written ...
The list of variables that are not currently saved in oo_, but should, and of results that should be stored in a way different from the current one.
Design Changes in Dynare 4.5
Implemented changes to the oo_-structure
The Bayesian forecasts stored in oo_.MeanForecast and oo_.PointForecast included the initial condition. Starting in Dynare 4.5, only the forecasts are stored in these fields. This makes their dimension consistent with classical forecasts.
The HPD fields generated for MCMC moments (HPDinf and HPDsup) were row vectors. Starting in Dynare 4.5 they will be column vectors like the other moment fields like Mean and Median.
Classical filtered variables stored in oo_.FilteredVariables were padded with zeros at the beginning and end of the vector. Starting in Dynare 4.5, the leading and trailing zeros are removed to make this field consistent with the Bayesian filtered variables.
The smoothed variables from classical estimation stored in SmoothedVariables now also store the smoothed variable including trend and constant. Before, it did not include trend and constant and thus was in deviations from the mean (in contrast to the Bayesian fields)
The trend component is stored in a Trend-subfield. This is necessary as the optimal trend can be estimated, but this estimate is nowhere reflected in the results.
- Constants and trends are also added to forecasts, to updated and to filtered variables.
The ordering in FilteredVariablesKStepAhead and FilteredVariablesKStepAheadVariances is now in declaration instead of decision rule ordering (see #943).
Proposed changes to the oo_-structure
In order to close ticket #339, results from classical estimation will be stored in oo_.ml. When doing this, the misleading naming of HPDIs will be changed to CI
The smoothed variables from Bayesian estimation stored in SmoothedVariables will store the smoothed variable including trend and constant. Currently the trend was not considered at all.
The convert_dyn_45_to_44 .m utility can be used to restore points 1-6 for backward compatibility reasons.
oo_.exo_simul is periods by exo_nbr, i.e. a collection of column vectors. In contrast, oo_.endo_simul is endo_nbr by periods , i.e. a collection of row vectors. This is confusing and leads to problems with initval-files where this needs to be taken into account. This should be harmonized in 4.5.