Thanks.
Does standard errors specified in the shocks block has to match the standard deviations of the priors of the shocks in estimation block?
No, those are unrelated objects.
Hello professor,
The only way for the estimation (with mode_compute=4) to work and provide decent results in my case is to add a measurement error term with a prior prescribed as below as below.
stderr w_obs, normal_pdf, 0, 1;
But the catch is I get a negative SD for the measurement term.
I hope you can give me some guidance and advise on how to navigate this issue.
Dynare internally works with variances, so the negative sign does not matter.
1 Like
In the NK_baseline_steadystate file there is a variable called zeta which I cannot find any reference to equations in the mod file or the steady state file. Is this a redundant variable?
Yes, it is. See NK_baseline_steadystate.m: cosmetic changes (!2282) · Merge requests · Dynare / dynare · GitLab
Hello professor,
In the FV(2010) paper, the author has transformed the sigma terms to their exponents as below.
However, in the results table, the estimated values are negative.
Since Inverse Gamma function is strictly positive, this cannot be. What priors would be suitable for a transformation like this?
The prior is for e^\sigma, which is strictly positive. But the posterior reported is for \sigma, which then ranges from [-\infty,+\infty]
Shouldn’t the standard deviations be positive theoretically?
To be precise, what is called \sigma in the paper is actually the log standard deviation \ln \sigma . So e^\sigma is the standard deviation.