#291: bugs related to auxiliary variables for lagged exogenous variables --------------------------+----------------------- Reporter: michel | Owner: sebastien Type: bug | Status: new Priority: major | Milestone: 5.0 Component: Preprocessor | Version: 4.3.1 Keywords: | --------------------------+----------------------- In ./tests/auxiliary_variables/test2.mod, the preprocessor fails to create auxiliary variable for lagged exogenous one.
When that is fixed, it will be necessary to check that the derivatives of the definition of the auxiliary variable is correctly added to test2_static.m
When test2.mod runs correctly, please add it to ./tests/Makefile.am
#291: bugs related to auxiliary variables for lagged exogenous variables ---------------------------+----------------------- Reporter: michel | Owner: sebastien Type: bug | Status: new Priority: major | Milestone: 5.0 Component: Preprocessor | Version: 4.3.1 Resolution: | Keywords: ---------------------------+-----------------------
Comment (by sebastien):
I am not sure this is a bug: since there is no computing task in this MOD file, the preprocessor does not know if we are in a deterministic or stochastic setup.
For deterministic setups, we don't create auxiliary variables for leaded/lagged exogenous variable. Looks like this is what is happening.
Can you try again after putting stoch_simul?
#291: bugs related to auxiliary variables for lagged exogenous variables ---------------------------+----------------------- Reporter: michel | Owner: sebastien Type: bug | Status: new Priority: major | Milestone: 5.0 Component: Preprocessor | Version: 4.3.1 Resolution: | Keywords: ---------------------------+-----------------------
Comment (by michel):
OK, then the test case as it is points to a bug in model_diagnostics when there are lags on more periods in the exogenous variables than in the endogenous ones. This problem may arise in other functions as well. We need additional tests for this case.
It doesn't work when stoch_simul is added because the Jacobian of the static model doesn't contains the derivatives of the definition of the auxiliary variable (that is an endogenous variable)
Note that a *.mod file with only steady is legitimate. It makes sense to start only with computing the steady state of a model and/or the user may want to use Dynare to do simply comparative statics. test2.mod works with only steady and no model_diagnostics.
#291: bugs related to auxiliary variables for lagged exogenous variables ---------------------------+----------------------- Reporter: michel | Owner: sebastien Type: bug | Status: new Priority: major | Milestone: 5.0 Component: Preprocessor | Version: 4.3.1 Resolution: | Keywords: ---------------------------+-----------------------
Comment (by sebastien):
Replying to [comment:2 michel]:
OK, then the test case as it is points to a bug in model_diagnostics
when there are lags on more periods in the exogenous variables than in the endogenous ones. This problem may arise in other functions as well. We need additional tests for this case.
It doesn't work when stoch_simul is added because the Jacobian of the
static model doesn't contains the derivatives of the definition of the auxiliary variable (that is an endogenous variable)
Then maybe we should rename that ticket to reflect the fact that it concerns model_diagnostics?
Note that a *.mod file with only steady is legitimate. It makes sense to
start only with computing the steady state of a model and/or the user may want to use Dynare to do simply comparative statics. test2.mod works with only steady and no model_diagnostics.
The preprocessor works when there is no computing task, but its behavior is not clearly defined. Currently, looking at the code, it seems it behaves like in deterministic mode (i.e. no jacobian wrt exogenous, no auxiliary variables wrt leaded/lagged exogenous). If this is not what is expected, a default behaviour should be made explicit.
#291: bugs related to auxiliary variables for lagged exogenous variables ---------------------------+----------------------- Reporter: michel | Owner: sebastien Type: bug | Status: new Priority: major | Milestone: 5.0 Component: Preprocessor | Version: 4.3.1 Resolution: | Keywords: ---------------------------+-----------------------
Comment (by michel):
I will open a ticket concerning model_diagnostics. However, test2.mod doesn't work when one adds stoch_simul for the reason stated in the description of the ticket. I will modify test2.mod accordingly.
By the way, steady should be considered as a computing statement, so should check and model_diagnostics.
A model without computing task is a model with only declarations. Such a model should produce a warning.
I think that the behavior described above when there is only steady is fine.
The different behavior for auxiliary variables for lagged/lead shock makes it impossible to have 1) a deterministic procedure 2) check 3) lead/lagged shocks
#291: bugs related to auxiliary variables for lagged exogenous variables ---------------------------+----------------------- Reporter: michel | Owner: sebastien Type: bug | Status: new Priority: major | Milestone: 5.0 Component: Preprocessor | Version: 4.3.1 Resolution: | Keywords: ---------------------------+-----------------------
Comment (by sebastien):
When I add stoch_simul in test2.mod, I get the derivatives of the auxiliary variable.
Can you try again, after verifying that your preprocessor binary is up-to- date?
#291: bugs related to auxiliary variables for lagged exogenous variables ---------------------------+----------------------- Reporter: michel | Owner: sebastien Type: bug | Status: closed Priority: major | Milestone: 5.0 Component: Preprocessor | Version: 4.3.1 Resolution: invalid | Keywords: ---------------------------+----------------------- Changes (by michel):
* status: new => closed * resolution: => invalid
Comment:
You are right this was due to a stale binary, I close the ticket.
-
Trac Dynare