Stéphane Adjemian authored 6 months ago and Stéphane Adjemian committed 3 months ago

[skip ci]

 + New routine get_user_defined_boundaries()

 + Add @dprior method for plotting prior densities.

 - check_mode_file() signature change (do not pass options_ as an
   argument, remove bayestopt_), bug fixes (wrong indexing of cell
   arrays) and add unit tests.

 + set of private functions under the estimation subfolder (code factorization).

 + initialguess() returns the initial guess for the estimated parameters

 + @dprior/isempty() method, return true iff object is empty

 + derror() Throws an error without the stack trace under
   MATLAB. Behaviour unchanged under Octave. Ideally,we should print
   the stack trace when in debug mode. Can we test if we are in debug
   mode without passing options_?

 + isbayesian() returns tre iff bayesian approach is considered.

 - set_prior returns a dprior object.

 - set_prior ne devrait en principe pas être appelé quand on ne
 considère pas une approche bayésienne... C'est pourtant le cas car
 cette fonction retourne aussi la condition initiale et les bornes de
 l'optimisateur. Il faut casser cette fonction.

uniform_specification() was throwing an error if the uniform prior is
defined with the expectation and variance.

For example, Prior.lb(3) returns the lower bound of the prior distribution for the third parameter, assuming Prior is a dprior object.

Indices for the parameters violating the prior boundaries.

Stéphane Adjemian authored 5 months ago and Stéphane Adjemian committed 3 months ago

 draw()  -> deviate()
 draws() -> sample()

Stéphane Adjemian authored 5 months ago and Stéphane Adjemian committed 3 months ago

Instead of exist().

Also rewrite some descriptions and add examples.

 - fill_
 - ge (>=)
 - gt (>)
 - le (<=)
 - lt (<)
 - hdiff, hdiff_
 - hgrowth, hgrowth_

🐛 subst_auxvar.m: account for auxiliary type 5

See merge request Dynare/dynare!2331

However do not set that option permanently (and also disable RemoveSemicolon by
default), since the clang-format manual states that these options can lead to
incorrect formatting and thus their result should be carefully reviewed.

Fixes crashes in model_info with differentiate_forward_vars

Preprocessor: fix performance regression on models with many nested model-local variables without leads/lags

Manual: document new syntax for Liu/West online filter

Closes #1948

See merge request Dynare/dynare!2332

Bug introduced in c4bf6d07.

Bug introduced in 095df542 and
9337ca95.

Bug introduced in 79128ea9.

In Sphinx 8, returned filesystem path objects are no longer extended to have
implicit string conversion methods. The conversion to a string must now be made
explicitly.

See this entry in the changelog for Sphinx 8.0:

 #12650, #12686, #12690: Extend the deprecation for string methods on
 :py:class:`~pathlib.Path` objects to Sphinx 9. Use :py:func:`os.fspath` to
 convert :py:class:`~pathlib.Path` objects to strings, or
 :py:class:`~pathlib.Path`'s methods to work with path objects. Patch by Adam
 Turner.

Closes #1948

🐛 Enable smoother_inversion_filter option with MCMC

See merge request !2325

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

[skip ci]

Previously, “contains('foo\', '\')” would fail with an error, due to the
antislash not being escaped when calling the “regexp” function. Same for other
special characters in regular expressions (e.g. the dollar sign).

In particular, this bug would make Dynare unusable with Octave under
Windows (since commit e9d79796), because
Windows paths typically contain antislashes.

Make compatible the outputs of the online particle filter with the Dynare...

See merge request Dynare/dynare!2323

As a consequence, drop workaround for https://github.com/msys2/MINGW-packages/issues/22712