Ricardo Vieira

You write on the social example: > What you can say for certain is that the true rate is less than your observed rate I assume you mean: > the...

Increase support for batched multivariate distributions

8

The `logp` of several multivariate distributions does not work (or is not tested) for arbitrarily batched dimensions. Some cases I could confirm include: - [ ] MvNormal - #5424 -...

Expand DensityDist docstrings

2

We should include code snippets about how to use the DensityDist, as 1) the API has changed since V3, and 2) it has always been a source of confusion https://github.com/pymc-devs/pymc/blob/7a6bd027ec827ee45056cfbe04b51b75643d2d19/pymc/distributions/distribution.py#L680-L686...

Raise informative error if dictionary is passed as observed of DensityDist

2

This is a common source of confusion in people switching from V3 to V4. We could print an informative message when users try to do it.

Otherwise they don't show up in the docs: https://www.pymc.io/projects/docs/en/v4.1.4/api/distributions/generated/pymc.DensityDist.html

Log name of RVs that are being resampled in posterior predictive

5

Right now, the behavior of `sample_posterior_predictive` is very opaque as to which RVs are being resampled. Any RVs that are downstream of Shared variables or other variables mentioned in `var_names`,...

Remove `samples` and `keep_size` from posterior_predictive

16

A bit more work than what was needed for #5772, mostly because we use `samples` in several tests. Quoting @OriolAbril > imo kill everything, `size`, `keep_size` and `samples` (to get...

Remove mypy from pre-commit

1

pre-commit is too slow since we introduced mypy in it, to the point I have disabled it and have to call it manually before pushing (if I remember to) I...

Follow up from #5438 Nested Mixtures or Mixtures using non pure RandomVariables (such as Censored variables) are not yet possible. This will require creating a dispatch to retrieve and manipulate...

Partially observed multivariate distributions not implemented in V4

2

I added an explicit `NotImplementedError` and future test in #5245 - https://github.com/pymc-devs/pymc/pull/5245/commits/022cefa3e6b64f73a8e91f927c4db3d7454c4418 Partial observed variables now rely on Aesara's `local_subtensor_rv_lift`, to separate the observed and unobserved portions of a RV...