hessian not positive definite and a laplace app. NaN

[tamara b]

Dear all,

I estimated a dsge model with a unit root using diffuse filter. The model works without the stimation part but reports several warning and error massages once the estimation part of the code is added.
First it gives a few warnings, that Jacobian is sigular and there is colinearity (but I guess this is fine given the UR), then this one
You did not declare endogenous variables after the estimation/calib_smoother command.

that I don't know how to understand,
then
POSTERIOR KERNEL OPTIMIZATION PROBLEM!
(minus) the hessian matrix at the "mode" is not positive definite!
=> posterior variance of the estimated parameters are not positive.
You should try to change the initial values of the parameters using
the estimated_params_init block, or use another optimization routine.
Warning: The results below are most likely wrong!
> In dynare_estimation_1 at 433
In dynare_estimation at 70
In teza_ss2_e at 519
In dynare at 120
Warning: Matrix is singular to working precision.
> In dynare_estimation_1 at 448
In dynare_estimation at 70
In teza_ss2_e at 519
In dynare at 120


and finally
Log data density [Laplace approximation] is NaN.

Error using chol
Matrix must be positive definite.


can you please help me understand to what error dynare might be pointing?

[jpfeifer]

1. Collinearity is to be expected when you have a unit root.
2.

You did not declare endogenous variables after the estimation/calib_smoother command.

just warns that after the estimation command, you did not specify a list of variables for which to compute endogenous moments. You can ignore this.

3. This is a common problem (dozens of posts in the forum). First of all, take a look at the mode check plots if there is anything strange (in particular if there are any horizontal lines indicating non-identifiability).

[tamara b]

thank you!
i digged a little through various topics and found a solution in setting the mode_compute to 6... and that solved all the issues..
now I have to look into how and why..