If the homogeneized equation evaluates to zero, then we skip the
test (otherwise the 2nd derivative of the log is infinite, and the test fails
while it should not necessarily).
Closes#506
Replace them by equivalents in M_ (and an extra one: M_.dynamic).
IMPORTANT POINT: oo_.dr.npred used to count both purely backward and mixed/both
variables. This was the cause of lots of confusion. The new M_.npred only
counts purely backward variables.
We now have the following indentities:
M_.npred + M_.nboth + M_.nfwrd + M_.nstatic = M_.endo_nbr
M_.nspred = M_.npred + M_.nboth
M_.nsfwrd = M_.nfwrd + M_.nboth
M_.ndynamic = M_.npred + M_.nboth + M_.nfwrd
+ the number of static, lag, lead and both variables,
+ the lead_lag_incidence matrix for variables related to previous blocks
+ the index of reordered policy rule
for each block during the preprocessing step
- correct a bug in the comment related to first order derivatives in _dynamic.m file for a block decomposed model without bytcode
- avoid simplification of the Jacobian matrix (setting to zero all elements below the cutoff) for estimated or stochastically simulated models: the cutoff option is set to 0.
- in the 3rd derivatives matrix, among symmetric elements, the first one had
the right values but the following ones were set to zero
- moreover, the k-order DLL was trying to add all the symmetric elements in the
folded tensor, instead of only keeping one value among all the symmetric ones
- hopefully, Ondra's tensor library was (silently) refusing to add symmetric
elements after the first (and right) value had been added
- so the final result was correct
and corrects the computation of the number of column in the jacobian matrices in a block decomposed model without bytecode
- Identical corrections for models using bytecode without block decomposition
- preprocessor provides informations about the dynamic structure of each block
- extends the algorithms available with bytecode: stack_solve_algo = 1, 2, 3 and 4 is compatible the bytecode. Speed and memory requirement are improved with stack_solve_algo = 1 or 4 for large scale models.
- bytecode can be used to evaluate a model
- remove useless output type oMatlabDynamicModelSparseLocalTemporaryTerms
(indistinguishable from oMatlabDynamicModelSparseLocal)
- fix bug in output type of sparse static model
* in DataTree, new functions for testing if some opcode (unary, binary or trinary) is used somewhere in the tree
* removed DataTree::containsSteadyState(), rather use a call to the new utility function
* fix options stack_solve_algo={2,3,4} (closes#68)
* fix crashes for singular normalizations (closes#44) and implement decreasing cutoff
* fail for stack_solve_algo=2 under Octave (because there is no gmres function in Octave)
git-svn-id: https://www.dynare.org/svn/dynare/trunk@3279 ac1d8469-bf42-47a9-8791-bf33cf982152
* fixed a bug with EXPECTATION operator: now handle model local variables
* substitute EXPECTATION operator even in deterministic models (for avoiding crashes, even if use of this operator doesn't make sense there)
* forbid the use of EXPECTATION and STEADY_STEATE outside model block
* minor implementation changes in EXPECTATION and STEADY_STATE operators
reference manual:
* documented EXPECTATION and STEADY_STATE operators
* added a note in write_latex_{static,dynamic}_model commands documentation about auxiliary variables and other model transformations
git-svn-id: https://www.dynare.org/svn/dynare/trunk@3145 ac1d8469-bf42-47a9-8791-bf33cf982152
* reworked a little bit the implementation
* documented the command in the reference manual
* added a test
git-svn-id: https://www.dynare.org/svn/dynare/trunk@3143 ac1d8469-bf42-47a9-8791-bf33cf982152
* fixed substitution of endogenous with leads >= 2: take into account exogenous with leads in non-linear terms
* fixed substitution of exogenous with leads: take into account other variables with leads in non-linear terms
git-svn-id: https://www.dynare.org/svn/dynare/trunk@3027 ac1d8469-bf42-47a9-8791-bf33cf982152
* In stochastic mode, now transforms the model by removing leads and lags greater or equal to 2 (creating auxiliary variables and equations in the process)
* Information about these variables is in structure M_.aux_vars
* Automatically add the necessary initialization for auxiliary vars after the initval block or load_params_and_steady_state
git-svn-id: https://www.dynare.org/svn/dynare/trunk@3002 ac1d8469-bf42-47a9-8791-bf33cf982152
* Added solve_algo=5 for steady
* Remove obsolete method option of simul
* Added various checks to avoid unimplemented combinations of bytecode/block/solve_algo/stack_solve_algo
git-svn-id: https://www.dynare.org/svn/dynare/trunk@2882 ac1d8469-bf42-47a9-8791-bf33cf982152
- the options "sparse_dll" and "sparse" are no more available. They are replaced by combinations of block (for block decomposition) and bytecode (for model stored in a binary file)
- markowitz is now an option of simul and steady commands
git-svn-id: https://www.dynare.org/svn/dynare/trunk@2873 ac1d8469-bf42-47a9-8791-bf33cf982152
- mfs: new option for 'steady' and 'model' commands. Determines the equation belonging to the set of feedback variables.
mfs = 0 => all variables are considered as feedback variables (default value)
mfs = 1 => using only naturally normalized equation as potential recursive equations (all variables assigned to unnormalized equations are considered as feedback variable)
mfs = 2 => adding to the set of potential recursive equation with mfs = 1 the linear equation in endogenous variable normalized (all variables assigned to nonlinear unnormalized equations are considered as feedback variable)
mfs = 3 => adding to the set of potential recursive equation with mfs = 2 the non linear equation in endogenous variable normalized
- correction of few buggs in simulate.dll
- block_mfs_dll: new option for 'steady' command. Use simulate.dll to solve the steady state model (speedup the computation of the steady-state and the homotopy)
git-svn-id: https://www.dynare.org/svn/dynare/trunk@2866 ac1d8469-bf42-47a9-8791-bf33cf982152
- no function comment in MinimumFeedbackSet.cc
- writeChaineRuleDerivative in DynamicModel.cc instead of ModelTree.cc
git-svn-id: https://www.dynare.org/svn/dynare/trunk@2836 ac1d8469-bf42-47a9-8791-bf33cf982152
* StaticModel: dropped creation of C file in USE_DLL mode
* StaticModel: added computation of derivatives for block MFS mode
* DynamicModel: catch error codes of mkdir()
* ModelTree: moved "mode" member variable to DynamicModel
* ModFile::writeOutput(): moved manipulations relative to dynamic model file to DynamicModel
git-svn-id: https://www.dynare.org/svn/dynare/trunk@2818 ac1d8469-bf42-47a9-8791-bf33cf982152
* moved some declarations from ExprNode.hh to BlockTriangular.hh to which they conceptually belong
* replaced non standard boolean operators (and, or, not) by standard ones (&&, ||, !)
git-svn-id: https://www.dynare.org/svn/dynare/trunk@2812 ac1d8469-bf42-47a9-8791-bf33cf982152
* fixed a bug in the handling of VariableNode: we now make a copy of the recursive_variables map, instead of modifying that of the caller
* factorized code shared with standard derivation
* various minor cleanups
git-svn-id: https://www.dynare.org/svn/dynare/trunk@2811 ac1d8469-bf42-47a9-8791-bf33cf982152
* for 2nd and 3rd derivatives of static and dynamic model, create the sparse matrices in a more efficient way (thanks to Pablo Winant for suggesting this and providing a patch)
* this breaks USE_DLL option at order 2
* fixed bug when hessian or 3rd deriv. matrix was all zero: the matrix was not constructed at all, leading to crashes in Matlab code
git-svn-id: https://www.dynare.org/svn/dynare/trunk@2787 ac1d8469-bf42-47a9-8791-bf33cf982152
- normalize an equation linear in its endogenous variable
- Chained rule derivatives (necessary to reduce a block to the feedback equations and variables)
git-svn-id: https://www.dynare.org/svn/dynare/trunk@2726 ac1d8469-bf42-47a9-8791-bf33cf982152
- First draft of DynamicModel.cc files with feedback variables.
TODO :
- reduction of the Jacobian matrix
- symbolic normalization of equations
- application to the binary code evaluation (simulate.dll).
git-svn-id: https://www.dynare.org/svn/dynare/trunk@2678 ac1d8469-bf42-47a9-8791-bf33cf982152
* added support for derivatives of Hessian w.r. to parameters
* added "identification" command which does nothing for the moment, except triggering those derivatives w.r. to params
git-svn-id: https://www.dynare.org/svn/dynare/trunk@2613 ac1d8469-bf42-47a9-8791-bf33cf982152
* fixed bug with planner_objective
* fixed another bug due to unitialized max_{lead,lags} values
git-svn-id: https://www.dynare.org/svn/dynare/trunk@2604 ac1d8469-bf42-47a9-8791-bf33cf982152
* created a distinct expression tree for the static model (thus giving better sharing of sub-expressions and better computation of temporary terms for the static model)
* for that purpose, created StaticModel and DynamicModel classes (ModelTree still persists, but only contains code shared between StaticModel and DynamicModel)
* removed sparse static file (to be later replaced by new algorithm for steady state computation on large models)
git-svn-id: https://www.dynare.org/svn/dynare/trunk@2592 ac1d8469-bf42-47a9-8791-bf33cf982152
Sébastien Villemot
Stéphane Adjemian (Charybdis)
Houtan Bastani
Johannes Pfeifer
Michel Juillard
Ferhat Mihoubi
Sébastien Villemot
Houtan Bastani
Michel Juillard
Houtan Bastani
ferhat
houtanb
sebastien
stepan
michel