Questions on estimation

Dynare help (legacy posts)
1

Hi professor!
I lately wrote a model and tried to estimate it. But there is a problem. Dynare can only give me posterior modes, after which It reported the following error message:
Error using chol
Matrix must be positive definite.

Error in metropolis_hastings_initialization (line 68)
d = chol(vv);

Error in random_walk_metropolis_hastings (line 62)
ix2, ilogpo2, ModelName, MetropolisFolder, fblck, fline, npar, nblck, nruns, NewFile, MAX_nruns, d ] = …

Error in dynare_estimation_1 (line 782)
feval(options_.posterior_sampling_method,objective_function,options_.proposal_distribution,xparam1,invhess,bounds,dataset_,options_,M_,estim_params_,bayestopt_,oo_);

Error in dynare_estimation (line 89)
dynare_estimation_1(var_list,dname);

Error in landfinance_bys_fullshock (line 713)
dynare_estimation(var_list_);

Error in dynare (line 180)
evalin(‘base’,fname) ;

And I couldn’t get posterior means if the problem is not solve. Does the matrix means covariance matrix for the proposal density of the MCMC sampler? I add an optionmcmc_jumping_covariance = prior_variance. But it said that my value of mh_init_scale was too large (0.4), and I had to change it to a very small value to let it proceed. The result is also weird. The condifence interval is too narrow that the upper value is almost equal to the lower value.

So…How can I modify my estimation option to get reliable posterior means and confidence interval?
My code is packed in zip file in the attachmentestimating.zip (11.2 KB). Thank you so much!

2

You need to correct your mod-file. First of all, check identification
You will see that

[quote]WARNING !!!
The rank of H (model) is deficient!

gltrans is not identified in the model!
[dJ/d(gltrans)=0 for all tau elements in the model solution!]
betaparam_trans is not identified in the model!
[dJ/d(betaparam_trans)=0 for all tau elements in the model solution!]
inveta is not identified in the model!
[dJ/d(inveta)=0 for all tau elements in the model solution!]

[/quote]

because you are trying to estimate parameters that are not even present in the data. Secondly, there is stochastic singularity in your model as there seems to be an exact linear combination between your observables implied by the model. You need to fix this as well.

After that, have a look at the mode_check
plots.