# Propagate posterior uncertainty as measurement error?

**URL:** <https://discourse.mc-stan.org/t/propagate-posterior-uncertainty-as-measurement-error/34486>\
**Category:** Modeling\
**Tags:** brms, hierarchical-model\
**Created:** [March 13, 2024, 10:49pm UTC](https://discourse.mc-stan.org/t/propagate-posterior-uncertainty-as-measurement-error/34486 "2024-03-13T22:49:10Z")\
**Posts on this page:** 1\
**Showing post:** 2

<div class="post-metadata">

**Author:** ![martinmodrak](https://yyz2.discourse-cdn.com/flex030/user_avatar/discourse.mc-stan.org/martinmodrak/32/133_2.png) [@martinmodrak](https://discourse.mc-stan.org/u/martinmodrak)\
**Post date:** [March 18, 2024, 1:05pm UTC](https://discourse.mc-stan.org/t/propagate-posterior-uncertainty-as-measurement-error/34486/2 "2024-03-18T13:05:13Z")

</div>

> [@kthayashi](#):
>
> `bf(Y ~ mi(Xmean)) + bf(Xmean | mi(Xsd) ~ 1) + gaussian() + set_rescor(FALSE)`

This is not a measurement error model, this is a missing data model. Measurement error models are specified in `brms` with `me()`, not `mi()`.

See [Predictors with Measurement Error in brms Models — me • brms](http://paul-buerkner.github.io/brms/reference/me.html) for examples. I’d guess this will match your use case reasonably well.

There are a bunch of forum posts here e.g.: [Using posteriors as new priors](https://discourse.mc-stan.org/t/using-posteriors-as-new-priors/12745) discussing potential drawbacks of this approximation (compared to a full joint model), so consider yourself warned :-)

Best of luck with your project!

---

_[View the full topic](https://discourse.mc-stan.org/t/propagate-posterior-uncertainty-as-measurement-error/34486)._