Add weight_predictions function

[aloctavodia]

This function will take a list of idatas with posterior_predictive groups and a list of model weights (computed using az.compare or something else) and it will return a new inference data with a posterior_predictive group composed of weighted samples from the input idatas.

@zaxtax maybe we can focus on this and forgot about pm.sample_posterior_predictive_w

• Follows official PR format
• Includes new or updated tests to cover the new feature
• Code style correct (follows pylint and black guidelines)
• Changes are listed in changelog

---

📚 Documentation preview 📚: https://arviz--2147.org.readthedocs.build/en/2147/

[zaxtax]

I think this is a fantastic idea. I always found `pm.sample_posterior_predictive_w` awkward. One check that I think we need is something ensuring all models have identical observed data.

…

On Sat, 29 Oct 2022, 15:04 Osvaldo A Martin, ***@***.***> wrote: This function will take a list of idatas with posterior_predictive groups and a list of model weights (computed using az.compare or something else) and it will return a new inference data with a posterior_predictive group composed of weighted samples from the input idatas. @zaxtax <https://github.com/zaxtax> maybe we can focus on this and forgot about pm.sample_posterior_predictive_w - Follows official <https://github.com/arviz-devs/arviz/blob/main/CONTRIBUTING.md#pull-request-checklist> PR format - Includes a sample plot to visually illustrate the changes (only for plot-related functions) - New features are properly documented (with an example if appropriate)? - Includes new or updated tests to cover the new feature - Code style correct (follows pylint and black guidelines) - Changes are listed in changelog <https://github.com/arviz-devs/arviz/blob/main/CHANGELOG.md#v0xx-unreleased> ------------------------------ You can view, comment on, or merge this pull request online at: #2147 Commit Summary - b9a599b <b9a599b> add weight_predictions File Changes (1 file <https://github.com/arviz-devs/arviz/pull/2147/files>) - *M* arviz/stats/stats.py <https://github.com/arviz-devs/arviz/pull/2147/files#diff-f1da0a2c4f56d6fa32279f759e04b4b1057a1266dc36f94d22347e5111a03bdc> (24) Patch Links: - https://github.com/arviz-devs/arviz/pull/2147.patch - https://github.com/arviz-devs/arviz/pull/2147.diff — Reply to this email directly, view it on GitHub <#2147>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAACCUKXHQGE7MRCB3ZLM2DWFUOFVANCNFSM6AAAAAARRZAIRA> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[zaxtax]

Should we just remove sample_posterior_predictive_w or throw an error pointing people to this function?

…

On Sat, 29 Oct 2022, 16:16 Rob Zinkov, ***@***.***> wrote: I think this is a fantastic idea. I always found `pm.sample_posterior_predictive_w` awkward. One check that I think we need is something ensuring all models have identical observed data. On Sat, 29 Oct 2022, 15:04 Osvaldo A Martin, ***@***.***> wrote: > This function will take a list of idatas with posterior_predictive > groups and a list of model weights (computed using az.compare or > something else) and it will return a new inference data with a > posterior_predictive group composed of weighted samples from the input > idatas. > > @zaxtax <https://github.com/zaxtax> maybe we can focus on this and > forgot about pm.sample_posterior_predictive_w > > - Follows official > <https://github.com/arviz-devs/arviz/blob/main/CONTRIBUTING.md#pull-request-checklist> > PR format > - Includes a sample plot to visually illustrate the changes (only for > plot-related functions) > - New features are properly documented (with an example if > appropriate)? > - Includes new or updated tests to cover the new feature > - Code style correct (follows pylint and black guidelines) > - Changes are listed in changelog > <https://github.com/arviz-devs/arviz/blob/main/CHANGELOG.md#v0xx-unreleased> > > ------------------------------ > You can view, comment on, or merge this pull request online at: > > #2147 > Commit Summary > > - b9a599b > <b9a599b> > add weight_predictions > > File Changes > > (1 file <https://github.com/arviz-devs/arviz/pull/2147/files>) > > - *M* arviz/stats/stats.py > <https://github.com/arviz-devs/arviz/pull/2147/files#diff-f1da0a2c4f56d6fa32279f759e04b4b1057a1266dc36f94d22347e5111a03bdc> > (24) > > Patch Links: > > - https://github.com/arviz-devs/arviz/pull/2147.patch > - https://github.com/arviz-devs/arviz/pull/2147.diff > > — > Reply to this email directly, view it on GitHub > <#2147>, or unsubscribe > <https://github.com/notifications/unsubscribe-auth/AAACCUKXHQGE7MRCB3ZLM2DWFUOFVANCNFSM6AAAAAARRZAIRA> > . > You are receiving this because you were mentioned.Message ID: > ***@***.***> >

[aloctavodia]

glad you like it. yeah we are assuming a few things here, we should add some check. We may want to deprecate pm.sample_posterior_predictive_w and point people here. and also update/improve the example in PyMC-examples

[codecov]

Codecov Report

Merging #2147 (c16de55) into main (2d88638) will decrease coverage by 0.02%.
The diff coverage is 76.19%.

@@            Coverage Diff             @@
##             main    #2147      +/-   ##
==========================================
- Coverage   90.70%   90.67%   -0.03%     
==========================================
  Files         120      120              
  Lines       12647    12667      +20     
==========================================
+ Hits        11471    11486      +15     
- Misses       1176     1181       +5     

Impacted Files	Coverage Δ
arviz/stats/__init__.py	100.00% <ø> (ø)
arviz/stats/stats.py	95.25% <76.19%> (-0.62%)	⬇️

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[ahartikainen]

Couple of comments

[aloctavodia]

ready for review or merge, depending on how benevolent you are feeling today

[oussamadhaoui]

ppc_w= az.stats.weight_predictions([model0, model1, model2], weights)
I used this line and I received this error
ValueError: All the InferenceData objects must contain the posterior_predictivegroup

[aloctavodia]

Hi @oussamadhaoui,

did you check that model0, model1, and model2 have the posterior_predictive group?

also for the future, it is usually the best idea to open a new issue ticket instead of adding comments to merged PRs. Check this guide