Marco Ratto authored 7 years ago and Stéphane Adjemian committed 7 years ago

(cherry picked from commit e45acc11)

Marco Ratto authored 7 years ago and Stéphane Adjemian committed 7 years ago

(cherry picked from commit a7bc6414)

Johannes Pfeifer authored 7 years ago and Stéphane Adjemian committed 7 years ago

https://github.com/DynareTeam/dynare/commit/99dbc8c74d0ecf4c7a610c4a394fd85028cf2741 introduced a bug by disabling the resetting of the prior bounds used for the MCMC. The manual clearly states the bounds are only operational during mode-finding, but not during MCMC (and therefore prior sampling)
(cherry picked from commit 127a543b)

(cherry picked from commit 410f8bd0)

Johannes Pfeifer authored 7 years ago and Stéphane Adjemian committed 7 years ago

Comment: A user provided a file in which the Sylvester equation solved using doubling can evaluate to inf. When Doubling fails, one can try the Hessenberg-Schur algorithm.

(cherry picked from commit b91fa10f)

(cherry picked from commit 2f4b883f)

(cherry picked from commit 48b58394)

(cherry picked from commit bfe07d38)

Marco Ratto authored 7 years ago and Stéphane Adjemian committed 7 years ago

(cherry picked from commit b3bcd2c8)

Fixed conflicts in matlab/initial_condition_decomposition.m.

Marco Ratto authored 7 years ago and Stéphane Adjemian committed 7 years ago

(cherry picked from commit 5e7256da)

Fixed conflicts in matlab/graph_decomp_detail.m. Also fixed issues due to the transition from
char arrays to cell of row char arrays for the variable names.

Johannes Pfeifer authored 7 years ago and Stéphane Adjemian committed 7 years ago

The singularity branch did not correctly account for the switching number of observables

(cherry picked from commit cbc0cdfe)

Johannes Pfeifer authored 7 years ago and Stéphane Adjemian committed 7 years ago

(cherry picked from commit 8e73289f)

(cherry picked from commit c4f19586)

Marco Ratto authored 7 years ago and Stéphane Adjemian committed 7 years ago

(cherry picked from commit b9741548)

Marco Ratto authored 7 years ago and Stéphane Adjemian committed 7 years ago

fix to the Kitigawa transformation that allows to reduce the computing time of the likelihood in large models, with a lot of static variables, by 30-50%.
This fixes the bug in e97e5c34 that led to 2f9dc092
It is tested and completely fixes the issue highlighted in #1312.

(cherry picked from commit cf8213f7)

Fixes the bug introduced in bbdde462.

Johannes Pfeifer authored 7 years ago and Stéphane Adjemian committed 7 years ago

Prevents freezing if dialog has not been opened before

(cherry picked from commit 9fadf19b)

Johannes Pfeifer authored 7 years ago and Stéphane Adjemian committed 7 years ago

(cherry picked from commit ec3fb76a)

Fixed conflict. In branch 4.5 the modified line was not under the condition testing if the model has auxiliary variables, and the
name of the structure was correct. So it was not the same bug. I suppose that Dynare was eventually crashing with models without
auxiliary variables (when sending the *_set_auxiliary_variables routine to the remotes), but I did not test.

(cherry picked from commit 3a622738)

Stéphane Adjemian authored 7 years ago and Houtan Bastani committed 7 years ago

Revert "macroprocessor: interpret arrays passed on the command line as arrays instead of strings. closes #1578"

Commit 204d9cd0 fails with gcc 4.9 (I checked that it works flawlessly
with 6.x). The error message is:

MacroDriver.cc:63:24: error: 'class std::basic_string<char>' has no member named 'front'
         if (it->second.front() == '[' && it->second.back() == ']')
                        ^
MacroDriver.cc:63:53: error: 'class std::basic_string<char>' has no member named 'back'
         if (it->second.front() == '[' && it->second.back() == ']')

Ideally we should bump the version of gcc used in the build system, but I will not do that before the
next bug fix release. I suppose it is easier to replace the front() and back() methods.

(cherry picked from commit 4d89452e)

Johannes Pfeifer authored 7 years ago and Stéphane Adjemian committed 7 years ago

(cherry picked from commit ef797eec)

Johannes Pfeifer authored 7 years ago and Stéphane Adjemian committed 7 years ago

(cherry picked from commit 28b499af)

(cherry picked from commit ad0bff37)

Johannes Pfeifer authored 7 years ago and Stéphane Adjemian committed 7 years ago

Workaround for Octave bug https://savannah.gnu.org/bugs/?52568

(cherry picked from commit 99e4cb6e)

Johannes Pfeifer authored 7 years ago and Stéphane Adjemian committed 7 years ago

Otherwise estimation with missing observations and prefilter option will crash

(cherry picked from commit 92549e48)

macroprocessor: interpret arrays passed on the command line as arrays instead of strings. closes #1578

(cherry picked from commit 204d9cd0)

Marco Ratto authored 7 years ago and Stéphane Adjemian committed 7 years ago

- fix to the case when no plot is done when there are too few behavioral/non-behavioral runs [as it already happens when one of the two sets is empty]
- added more explicit messages on screen

(cherry picked from commit ec35210e)

Marco Ratto authored 7 years ago and Stéphane Adjemian committed 7 years ago

(cherry picked from commit f4b25efc)

Marco Ratto authored 7 years ago and Stéphane Adjemian committed 7 years ago

1) definition of annualized_realtime_conditional_shock_decomposition
2) making plots when vintage is not Q4. The latter required some restructuring and factorization of the code

(cherry picked from commit 3e589d24)

Fixed conflicts:
	matlab/annualized_shock_decomposition.m

Marco Ratto authored 7 years ago and Stéphane Adjemian committed 7 years ago

(cherry picked from commit fd8a69d9)

Marco Ratto authored 7 years ago and Stéphane Adjemian committed 7 years ago

- set xind_tick for very short time intervals in plots
- provisions for change in behavior of matlab bar function since R2017b

(cherry picked from commit a04b2876)

Marco Ratto authored 7 years ago and Stéphane Adjemian committed 7 years ago

- wrong output list of evaluate_smoother (M missing_)
- realtime option must be  forced to be zero
- wrong name of option fig_names

(cherry picked from commit 5d6b688f)

Marco Ratto authored 7 years ago and Stéphane Adjemian committed 7 years ago

(cherry picked from commit 45166aba)

Johannes Pfeifer authored 7 years ago and Houtan Bastani committed 7 years ago

Closes #1548

(cherry picked from commit ff1ae57e)

(cherry picked from commit d57b0723)

(cherry picked from commit 43d910a6)

Stéphane Adjemian authored 7 years ago and Stéphane Adjemian committed 7 years ago

Make isfile.m command work with Matlab versions before 9.1

(cherry picked from commit bb7c2450)