# Posterior Standard Errors

**URL:** <https://forum.dynare.org/t/posterior-standard-errors/351>\
**Category:** Dynare help (legacy posts)\
**Created:** [28 May 2007 17:34 UTC](https://forum.dynare.org/t/posterior-standard-errors/351 "2007-05-28T17:34:25Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![pkapinos](https://forum.dynare.org/letter_avatar_proxy/v4/letter/p/a183cd/32.png) [@pkapinos](https://forum.dynare.org/u/pkapinos)\
**Post date:** [28 May 2007 17:34 UTC](https://forum.dynare.org/t/posterior-standard-errors/351/1 "2007-05-28T17:34:25Z")

</div>

Hello,  
I’d really appreciate it if someone could tell me where the posterior standard errors are stored. When I type:

oo\_.posterior\_std.parameters

I get numbers that are very different from the confidence interval output that is reported automatically. Or do those confidence intervals not correspond to the posterior standard errors?

For instance, I have a parameter that follows the normal distribution. The point estimate is 0.79, the 95% confidence interval is 0.61-0.91 and the corresponging output for the posterior standard error is 0.81.

Thank you very much!  
Pavel

---

<div class="post-metadata">

**Author:** ![MichelJuillard](https://forum.dynare.org/user_avatar/forum.dynare.org/micheljuillard/32/5318_2.png) [@MichelJuillard](https://forum.dynare.org/u/MichelJuillard)\
**Post date:** [28 May 2007 18:28 UTC](https://forum.dynare.org/t/posterior-standard-errors/351/2 "2007-05-28T18:28:02Z")

</div>

Your parameter follows a normal prior, but we don’t know what if the shape of its posterior. That is why we need the Metropolis iterations for.  
The reported confidence interval is a 90% high probability region that isn’t directly related to the standard deviation of the posterior distribution.

Kind regards

Michel

---

<div class="post-metadata">

**Author:** ![pkapinos](https://forum.dynare.org/letter_avatar_proxy/v4/letter/p/a183cd/32.png) [@pkapinos](https://forum.dynare.org/u/pkapinos)\
**Post date:** [29 May 2007 01:26 UTC](https://forum.dynare.org/t/posterior-standard-errors/351/3 "2007-05-29T01:26:31Z")

</div>

Sorry, that was a pretty inadequate question; thank you for answering it nonetheless. I’d really appreciate it if you could suggest how to construct Bayesian IRF confidence intervals on my own, however. The mean response is stored in oo\_.PosteriorIRF.dsge.Mean. Does oo\_.PosteriorIRF.dsge.Var contain the variance of the impulse response? So that the reported bands are  
oo\_.PosteriorIRF.dsge.Mean.VarName±2\*sqrt(oo\_.PosteriorIRF.dsge.Var.VarName)? Thanks again.