Update documentation on how ess is handled

stan-dev / posteriordb

Database with posteriors of interest for Bayesian inference

176 stars 36 forks source link

Update documentation on how ess is handled #122

Closed MansMeg closed 3 years ago

MansMeg commented 4 years ago

We now check that the ESS is roughly within 4 SD from a NULL distribution (that is roughly sqrt(7 * S) where S is the total number of draws. This is a rough check for the ESS values, but far from perfect.

In the long run, we want to check this with a hierarchical model instead.

avehtari commented 4 years ago

Be more specific. ESS for which quantity?

MansMeg commented 4 years ago

Sure. ESS_bulk and ESS_tail for all parameters.

MansMeg commented 3 years ago

Now done.