I get strange red dots on the horizontal (“x”) axes of several of the marginal likelihood/posterior plots. Are these true zeros? Or are they, as a suspect, warnings of NaN and Inf?

Thanks for the help,
Rob Luginbuhl

p.s. anyone want to respond to my earlier post about this problem:

{??? Error using ==> schur
Input to SCHUR must not contain NaN or Inf.

mode_check displays red bullets on the horizontal axis for values of the parameter violating the Blanchard and Kahn conditions (also for values such that Dynare fails in computing the deterministic steady state).

For the other question I don’t know, I do not have enough information here.

I think that I get a similar problem as Rob and I would be interested in understanding what is happenning exactly.
Here are my mod file, my steady state file. I have not been able to upload my mat file neither, so I upload the raw data in .txt format and the matlab file to transform them .
Thanks

mode_check displays red bullets on the horizontal axis for values of the parameter violating the Blanchard and Kahn conditions (also for values such that Dynare fails in computing the deterministic steady state).

[/quote]

Will the estimated_params_bound command help to eliminate,for the parameters with red dots, the interval under which there is no stable solution to the system by not permitting the parameter to go there?

This is not likely… The estimated_params_bound block is only used to
constrain (on a specified interval) the optimizer when estimating the posterior mode…

The mode_check command triggers the evaluation of the posterior kernel
and likelihood around the estimate of the posterior mode. By default
the size of the neighborhood is +/- 50% of the posterior mode
estimate. For instance, if the posterior mode of a parameter is found
to be equal to .5, then Dynare will evaluate the likelihood and
posterior kernel for values of this parameter between .25 and
.75. Obviously this interval is very often too large. You can reduce
the size of the neighborhood around the estimated posterior mode by
writting something like:

options_.mode_check_neighbourhood_size = 0.05;

before the estimation (for a +/- five percent interval
around the estimate).