Re: [stan-dev/stan] Uniform sampling of finitely-bounded variable in stan 2.10 (#1972)

[Bob Carpenter]

[moved off issue tracker]

Thanks for the description. I just didn't know it by name. It's the
method is what Cook, Gelman and Rubin use.

http://andrewgelman.com/2013/05/23/validation-of-software-for-bayesian-models-using-posterior-quantiles-2/

- Bob

> On Jul 26, 2016, at 6:46 PM, Avraham Adler <notifi...@github.com> wrote:
>
> See probability integral transform. The quantiles of set of randomly generated observations should be uniformly distributed. So regardless of the generating distribution, if you convert the observations to the corresponding points on the CDF through F^{-1}(X) the resulting values should be uniformly distributed between 0 and 1. It's the inverse of using the CDF to generate observations.
>
> —
> You are receiving this because you were mentioned.
> Reply to this email directly, view it on GitHub, or mute the thread.
>

[Avraham Adler]

Good to know; thanks for the link.

Avi