Houtan Bastani
e3c12f5094
var_expectation: stop processing if variable used that was not in model
2017-05-04 16:56:50 +02:00
Houtan Bastani
d4376c1dfe
var_model: fix comments, copyright
2017-05-04 16:32:37 +02:00
Houtan Bastani
63180c9e59
var_model: allow non model-block variables in VAR
2017-05-04 16:15:33 +02:00
Stéphane Adjemian (Charybdis)
d0071bd80b
Fixed subsample periods for exogenous variables.
...
Model inversion routine was returning wrong results when the periods in
constrainedpaths where not aligned with the periods in exogenousvariables.
2017-05-03 14:06:38 +02:00
Stéphane Adjemian (Charybdis)
c2556cc80e
Merge branch 'master' into ecb-master
...
Model inversion of backward looking models.
2017-05-02 23:15:42 +02:00
Stéphane Adjemian (Charybdis)
5ce20179bd
Added specialized code for the inversion of backward models.
2017-05-02 23:04:44 +02:00
Stéphane Adjemian
f6bb392165
Merge pull request #1445 from JohannesPfeifer/ramsey_message
...
Ramsey: provide informative error messages related to steady state
2017-05-02 22:56:26 +02:00
Johannes Pfeifer
37bffddef8
Provide correct error code if no steady_state-file is provided and Ramsey fails to find steady state
2017-05-02 13:48:25 +02:00
Johannes Pfeifer
5af9af06b4
Provide informative error message if steady state file is used without specifying an instrument
2017-05-01 18:59:11 +02:00
Stéphane Adjemian (Charybdis)
e1c983cc7b
Fixed bug (ensure that we concatenate column vectors).
2017-04-29 19:20:02 +02:00
Johannes Pfeifer
b52c095a6b
Add legacy function for loading mat-files with old interface
...
Follows approach of #758 and closes #1364
2017-04-29 18:25:39 +02:00
Stéphane Adjemian
072246c5ab
Merge pull request #1441 from JohannesPfeifer/ksstat
...
Remove redundant ksstat option
2017-04-29 18:00:25 +02:00
Johannes Pfeifer
d7b2e271d7
Update documentation on sensitivity
...
Closes #789
2017-04-27 12:03:02 +02:00
Johannes Pfeifer
451646ff16
Remove redundant ksstat option
...
Closes #1431
2017-04-27 11:08:17 +02:00
Stéphane Adjemian (Charybdis)
c1b6a58eb7
Manually revert 05fc096569
.
...
Robust prediction error covariance matrix computation is now optional (with
rescale_prediction_error_covariance option).
Closes #1437 .
2017-04-27 10:44:27 +02:00
Stéphane Adjemian (Charybdis)
d6e1d0baa5
Added option rescale_prediction_error_covariance.
2017-04-26 18:22:03 +02:00
Stéphane Adjemian (Charybdis)
e21387c874
Bump Octave version.
...
Octave 4.2.1 32bit or 64bit is now required (should work with 4.2.0 but not
tested anymore, so users should upgrade).
2017-04-26 15:51:24 +02:00
Stéphane Adjemian (Charybdis)
650a3f6da2
Point to mex/octave32 if OCtave 32bit is used on Windows platform.
2017-04-26 15:10:40 +02:00
Johannes Pfeifer
6ca415ce14
stochastic_solvers.m: Suppress redundant warning if order was already set to 1
2017-04-25 20:16:10 +02:00
Houtan Bastani
d8ce2e051d
bug: var_expectation: account for cases when a variable is not present in a period in M_.lead_lag_incidence
2017-04-24 14:34:48 +02:00
Houtan Bastani
7829a25238
make stoch_simul solve linear model using order=1, even when order=2,3 specified. Also makes stoch_simul more efficient. closes #1336
...
(cherry picked from commit e29be9676dd40388919819af2d366b8308763ccc)
2017-04-22 16:26:13 +02:00
Stéphane Adjemian (Charybdis)
767e45db95
Shut off warning about short circuit operators (when calling qzdiv).
2017-04-22 14:44:47 +02:00
Stéphane Adjemian (Charybdis)
ad21612967
Removed Octave specific part in mjdgges.m function.
...
Also force complex QZ factorization in Octave.
2017-04-22 14:44:47 +02:00
Michel Juillard
b5d131bfa6
VAR: fixed bug where length() is used on character matrix
2017-04-22 10:11:17 +02:00
Johannes Pfeifer
fa1563370c
Make resol.m honor nocheck flag
...
Before, we never tested for correctness, allowing for wrong steady state files in the check command. Fixes bug introduced in 616efb53a448cfa59149951395838627fd1bf719 and restores behavior in 4.4.3
2017-04-21 10:10:03 +02:00
Stéphane Adjemian
7f4c1c6529
Merge pull request #1435 from JohannesPfeifer/cmaes_fitness
...
mode_compute=9: Make sure initial search volume (SIGMA) is not badly …
2017-04-20 16:33:37 +02:00
Stéphane Adjemian (Charybdis)
7376498e51
Fixed branch for dseries submodule.
2017-04-20 16:28:39 +02:00
Johannes Pfeifer
52a082a1f9
Add check for positive definiteness of Sigma_u^star
...
Otherwise, complex likelihoods may occur
2017-04-20 16:18:03 +02:00
Stéphane Adjemian (Charybdis)
d0682bca30
Updated dseries submodule.
2017-04-20 15:36:06 +02:00
Stéphane Adjemian (Charybdis)
27bffd27fc
Updated dseries submodule.
2017-04-20 15:33:57 +02:00
Houtan Bastani
db535236b2
Merge branch 'master' into ecb-master
2017-04-14 11:35:43 +02:00
Stéphane Adjemian (Charybdis)
ead9982d91
Updated dseries submodule (fix for passing NaNs to FAME database).
2017-04-14 11:18:35 +02:00
Johannes Pfeifer
f1123200e5
mode_compute=9: Make sure initial search volume (SIGMA) is not badly conditioned
2017-04-12 13:39:34 +02:00
Stéphane Adjemian (Charybdis)
97ac502d3d
Fixed wrong threshold (unit root detection again).
2017-04-05 14:14:56 +02:00
Stéphane Adjemian (Charybdis)
288f010944
Fixed bug (detection of unit roots).
2017-04-05 14:11:23 +02:00
Stéphane Adjemian (Charybdis)
f57fc0a5a3
Added new routine for model inversion.
2017-04-05 13:42:03 +02:00
Stéphane Adjemian (Charybdis)
8a18e87d98
Removed persistent variables in evaluate_smoother routine.
...
Closes #1412 .
2017-04-05 11:44:27 +02:00
Johannes Pfeifer
00299a92c1
Make sure options_.first_obs is properly set
2017-04-05 11:44:27 +02:00
Stéphane Adjemian (Charybdis)
4dcdb81a31
Changed error message introduced in 30c9b3721564d89d1b85abcdc75d493075e6ca77.
2017-04-03 11:25:05 +02:00
Stéphane Adjemian (Charybdis)
ade89a9ec1
Fixed missing initialization of qz_criterium with nonlinear filters.
2017-04-03 11:25:05 +02:00
Johannes Pfeifer
76c4bd04fe
Provide informative error message if non-stationary model is estimated using particle filter
...
Closes #1377
(cherry picked from commit efaa7e54136ffd48fb69e075ecacf9708a553d3b)
2017-04-03 11:25:05 +02:00
Stéphane Adjemian (Charybdis)
f9a462bf07
Added option nonlinear_filter_initialization.
...
Default value is 1 (initialization with the ergodic variance of the reduced
form solution of the model approximated at order one).
If the model has unit roots, the user must use `nonlinear_filter_initialization=3`,
which select an identity matrix for the initial covariance matrix of the state variables.
A side effect of this option is to temporarily change the value of options_.qz_criterium to
a value above one (ie 1+1e-6) so that the unit roots are not rejected. If the
model has unit roots and if the and if the option
nonlinear_filter_initialization has a value less than 3, the evaluation of the
likelihood will fail, because by default the unit root is counted as an
unstable root.
2017-04-03 11:25:05 +02:00
Stéphane Adjemian (Charybdis)
97e185bc6e
Cosmetic changes in doc header.
2017-04-03 11:25:05 +02:00
Houtan Bastani
f18c16653e
Merge pull request #1428 from JohannesPfeifer/patch-1
...
Test the rank of Pinf in univariate smoother for the subset of observables
2017-04-01 09:57:32 +02:00
Johannes Pfeifer
1755192e63
Make 792924 consistent with cases where Pinf is an empty matrix
...
Fixes the crash in `analytic_derivatives/fs2000_analytic_derivation.mod`
2017-03-31 17:41:38 +02:00
Houtan Bastani
ac851b0bac
preprocessor: initial_condition_decomposition front end. closes #1425
2017-03-31 13:19:04 +02:00
Houtan Bastani
3e779ebfee
initial_condition_decomposition: fix options_ substructure
2017-03-31 13:09:21 +02:00
Marco Ratto
792924a3b7
- test the rank of Pinf in univariate smoother for the subset of observables, in order to trap cases where this is not already trapped in compute_Pinf_Pstar.m
...
- issue warning when diffuse steps never end [maybe we should issue an error?]
2017-03-30 11:21:54 +02:00
Johannes Pfeifer
bf607ce630
Make sure no stale _posterior_draws.mat files are present
...
May crash compute_moments_varendo.m due to more files being present than generated in the current run
2017-03-29 14:52:32 +02:00
Johannes Pfeifer
839e7075f6
dsge_simulated_theoretical_covariance.m: fix bug that led to last _2ndOrderMoments*.mat file not being saved, instead overwriting the prior one
...
Produced wrong posterior covariance estimates in large models, because the moments matrix contained zeros that were not overwritten by loaded draws due to missing file
2017-03-29 14:52:32 +02:00