Problems with the identification command

[fabiac]

Hi there,

I had three problems with the global sensitivity routines and would appreciate if you could tell me what's going on. Just for clarification, I downloaded the proper GSA toolbox for each version of Dynare I'm using:

Problem #1:
Using Dynare 4.2.1 and Matlab R2009a or Matlab R2011a, if I run the demo file in the GSA toolbox, ls2003ide.mod, I get the following error messages:

??? Attempted to access aux(:,2); index out of bounds
because size(aux)=[0,0].

Error in ==> map_ident_ at 100
k = find(aux(:,2) > npred);

Error in ==> dynare_sensitivity at 201
map_ident_(OutputDirectoryName);

Error in ==> ls2003ide at 193
dynare_sensitivity(options_gsa);

Error in ==> dynare at 132
evalin('base',fname) ;


Problem #2:
Using Dynare 4.1.3 and Matlab R2009a, if I run the demo file in the GSA toolbox, ls2003ide.mod, I get the following error:

??? Undefined function or method 'gsa_sdp' for input
arguments of type 'double'.

Error in ==> map_ident_ at 1001
gsa_(j) = gsa_sdp(y0(1:nest), x0(1:nest,imap), ...

Error in ==> dynare_sensitivity at 189
map_ident_(OutputDirectoryName);

Error in ==> ls2003ide at 201
dynare_sensitivity(options_gsa);

Error in ==> dynare at 132
evalin('base',fname) ;



Problem #3:
Using Dynare 4.1.3 and Matlab R2009a, if I change the ls2003ide.mod file to include the command "identification;" in substitution to all those lines that refer to the global_sensitivity analysis, I get a number of error messages:

Warning: Divide by zero.
> In dynare_identification at 254
In ls2003ide_teste at 184
In dynare at 132
Parameter e_q is not identified in the model for 100% of MC runs!

Parameter e_q is not identified by J moments for 100% of MC runs!

Parameter e_A is not identified in the model for 100% of MC runs!

Parameter e_A is not identified by J moments for 100% of MC runs!

Parameter e_ys is not identified in the model for 100% of MC runs!

Parameter e_ys is not identified by J moments for 100% of MC runs!

Parameter e_pies is not identified in the model for 100% of MC runs!

Parameter e_pies is not identified by J moments for 100% of MC runs!

Parameter psi1 is not identified in the model for 100% of MC runs!

Parameter psi1 is not identified by J moments for 100% of MC runs!

Parameter psi2 is not identified in the model for 100% of MC runs!

Parameter psi2 is not identified by J moments for 100% of MC runs!

Parameter psi3 is not identified in the model for 100% of MC runs!

Parameter psi3 is not identified by J moments for 100% of MC runs!

Parameter rho_R is not identified in the model for 100% of MC runs!

Parameter rho_R is not identified by J moments for 100% of MC runs!

Parameter alpha is not identified in the model for 100% of MC runs!

Parameter alpha is not identified by J moments for 100% of MC runs!

Parameter rr is not identified in the model for 100% of MC runs!

Parameter rr is not identified by J moments for 100% of MC runs!

Parameter k is not identified in the model for 100% of MC runs!

Parameter k is not identified by J moments for 100% of MC runs!

Parameter tau is not identified in the model for 100% of MC runs!

Parameter tau is not identified by J moments for 100% of MC runs!

Parameter rho_q is not identified in the model for 100% of MC runs!

Parameter rho_q is not identified by J moments for 100% of MC runs!

Parameter rho_A is not identified in the model for 100% of MC runs!

Parameter rho_A is not identified by J moments for 100% of MC runs!

Parameter rho_ys is not identified in the model for 100% of MC runs!

Parameter rho_ys is not identified by J moments for 100% of MC runs!

Parameter rho_pies is not identified in the model for 100% of MC runs!

Parameter rho_pies is not identified by J moments for 100% of MC runs!



Multi collinearity in the model:

param min mean max
e_R 0.000000 0.000000 0.000000


Multi collinearity for moments in J:

param min mean max
e_R 0.000000 0.000000 0.000000
Warning: Divide by zero.
> In myboxplot at 102
In dynare_identification at 455
In ls2003ide_teste at 184
In dynare at 132
??? Error using ==> set
Bad property value found.
Object Name : axes
Property Name : 'YLim'
Values must be increasing and non-NaN.

Error in ==> myboxplot at 144
set(gca, 'YLim', [mm-(MM-mm)*0.05-eps,
MM+(MM-mm)*0.05+eps]);

Error in ==> dynare_identification at 455
myboxplot(squeeze(idelre.Pco(immm(2:nbox+1),j,:))'),

Error in ==> ls2003ide_teste at 184
dynare_identification(options_ident);

Error in ==> dynare at 132
evalin('base',fname) ;

[rattoma]

Hi,

I will check and fix these errors and let you know.
Thank you for pointing to these errors
best
Marco

[rattoma]

Concerning problem #2, the mapping toolbox is missing there
Concerning problem #3, there may be a bug there, but I advise to use more recent versions of dynare. There will not be new releases of dynare 4.1.
I am fixing problem #1 and release a debugged version of the gsa toolbox for version 4.2.2

[fabiac]

Thanks a lot!