Johannes Pfeifer authored 4 months ago and Sébastien Villemot committed 4 months ago

Closes #1945

It has been available in Octave for a long time.

This will facilitate the introduction of dates in shock sequences.

Field names used in the option structure were slightly different depending on
whether the option value was an integer or a date. This is confusing.

This commit aligns the structure field names with the option names. It also
simplifies the logic for histval_file/initval_file option handling.

[skip ci]

Ref. #1859

Ref. #1859

The legacy representation is no longer used for the estimation chain.

Note that this mechanism is not robust (in particular it does not work with the
use_dll option), but this is not new.

Ref. #1859

Ref. #1859

Bytecode does not support conditional forecasting with block
decomposition (since the derivatives w.r.t. exogenous are no longer provided by
the preprocessor when doing block decomposition). In practice trying this would
lead to a crash.

Incidentally remove calling store_smoother_results twice

It comes after a “return” statement.

Keep only the syntax with 3 arguments: plan, dataset, range.

Removes unused prior_analysis_endo_var_list option

GetAllPosteriorDraws.m: make routine more efficient by enabling to output full parameters matrix instead of column vectors

Also factorizes loading of record file to obtain MCMC info by moving it into GetAllPosteriorDraws.m

Reverts of Dynare/dynare@5424fde7

moments_varendo and prior_posterior_statistics: account for different paths of various posterior samplers

Incidentally removes persistent variables used in metropolis_draw.m

Incidentally:
– fix a bug where path for endogenous after failure to converge was not centered
– ensure that sim1_linear.m is actually tested, by adding a test for
  linear_approximation + stack_solve_algo=0

Ref. #1859

⇒ Generated pseudo-random numbers will not change if we add new unit tests.

Ref. #1859

Ref. #1859

One input argument is required