Page 1 of 1

Conf.Interval of prior

PostPosted: Mon Feb 13, 2017 1:03 pm
by KKLS
Hi all,

I am afraid this is too basic to ask. I couldn ot figure it.

In bayesian estimation , If I set a PRIOR for a parameter as

\rho, BETA_PDF, a, b;


Is there a way I can also calculate the confidence interval for this prior ?


I know it might be basic,

Thanks in advance ?

Re: Conf.Interval of prior

PostPosted: Mon Feb 13, 2017 1:19 pm
by jpfeifer
Confidence intervals and Bayesian econometrics do not square well. Bayesian look at credible sets. What exactly do you mean? The second number in the specified prior is the standard deviation of the prior, which gives you an indiciation of the prior precision.

Re: Conf.Interval of prior

PostPosted: Mon Feb 13, 2017 1:56 pm
by KKLS
what is the credibel set (interval ) that this particular prior defines ?
Can I calculate it or is dynare calculating the posterior credible set and that should also be the prior credible set (is it the same ?)

in one of the papers it says :
I am setting this particular prior so that the 95 % credible set is this [ m;n] (some values) ..

many thankx Professor

Re: Conf.Interval of prior

PostPosted: Mon Feb 13, 2017 3:40 pm
by jpfeifer
At https://github.com/JohannesPfeifer/prior_from_quantiles you can find codes to compute quantiles for various distributions. See Koop's textbook Bayesian Econometrics for a definition of a credible set.

Re: Conf.Interval of prior

PostPosted: Mon Feb 13, 2017 5:09 pm
by KKLS
Thanks for the reply Professor.

1. I was trying to say is there a way I can calculatete the credible set given the PRIOR Median and S.E ?

2. To do the reverse :

I just copy pasted the 4 codes files .

- run_prior_parameter_calculation.m
- get_gamma_ab.m
- get_inv_gamma_ab.m
- cmaes.m
and ran the first one (run_prior_parameter_calculation.m) Prof. Pfeifer.

I got such an error :

run_prior_parameter_calculation
Error using feval
Undefined function 'get_gamma_ab' for input arguments of type 'double'.

Error in cmaes (line 691)
fitness.hist(1)=feval(fitfun, xmean, varargin{:});

Error in run_prior_parameter_calculation (line 50)
[~,~,~,~,~,bestever] = cmaes('get_gamma_ab',x_start,0.2,[],target);


Wonder if I did sth wrong ?!!

Re: Conf.Interval of prior

PostPosted: Mon Feb 13, 2017 6:14 pm
by jpfeifer
1. The easiest way is to use the unstable version together with
Code: Select all
write_latex_prior_table

to generate a LaTeX table or use the command line
Code: Select all
prior table

Both require a previous
Code: Select all
estimated_params
block. See the manual for details.

2. Check whether the mentioned file is really in the current folder.

Re: Conf.Interval of prior

PostPosted: Thu Feb 16, 2017 2:43 pm
by KKLS
Many Thanks Prof. Pfeifer,

In ''dynare.pdf'' (of the unstable version) I can find the documentation for the first command
write_latex_prior_table;


but NOT for the second one, i.e.
prior table


Anywhere else I can check for it pls ?

Re: Conf.Interval of prior

PostPosted: Thu Feb 16, 2017 3:03 pm
by jpfeifer
You should find it at the bottom of Section "9 Dynare misc commands"

Re: Conf.Interval of prior

PostPosted: Tue Feb 21, 2017 3:47 pm
by KKLS
Many Thanks Professor!