Updated multilevel modeling to v4 #285

fonnesbeck · 2022-02-27T03:57:31Z

Fixed models to work in v4.0.0b2 and added missing remote images as local files.

I don't like the new formulation of the early models as having independent means for floor and basement, but I can tackle that issue in a separate PR.

This should supersede #282

review-notebook-app · 2022-02-27T03:57:35Z

Check out this pull request on

See visual diffs & provide feedback on Jupyter Notebooks.

Powered by ReviewNB

review-notebook-app · 2022-02-27T22:00:00Z

View / edit / reply to this conversation on ReviewNB

OriolAbril commented on 2022-02-27T21:59:59Z
----------------------------------------------------------------

The date should be updated too.

review-notebook-app · 2022-02-27T22:00:00Z

View / edit / reply to this conversation on ReviewNB

OriolAbril commented on 2022-02-27T22:00:00Z
----------------------------------------------------------------

Line #1.    coords = {"Level": ["Basement", "Floor"], "obs_id": np.arange(floor.size)}

There is no need anymore for "decoy" coordinates. Using obs_id for the first time in floor_idx will define it. The default coordinate values are an arange, so the output should not change. It should also use the mutable argument or one of the Constant/Mutable data classes to get rid of the warning.

fonnesbeck commented on 2022-02-27T22:29:12Z
----------------------------------------------------------------

Sorry, not sure what you mean by decoy coordinates. Do they not have to be in the coords dict to be used by floor_idx?

OriolAbril commented on 2022-02-27T22:47:13Z
----------------------------------------------------------------

no, floor_idx is already an array of the right shape, so dims here have only 2 functions (out of 3 possible functions): annotating the dimension, and checking the shape matches for all objects annotated with that dimension. It is not being used to define the shape like it would in pm.Normal("var", 0, 1, dims="obs_id").

With the improved Aesara capabilities and the refactors to shape/size/dims functionality done by Brandon and Michael (I think it was mostly the two of them) when annotating dimensions, we can use dimension names that are not in coords which is very convenient when used in conjunction with pm.Data because the dimension and the variable where it is first annotated are linked so when generating predictions we only need to update the data instead of both data and coordinate values. See for example https://github.com/pymc-devs/pymc/blob/main/pymc/tests/test_shape_handling.py#L336

review-notebook-app · 2022-02-27T22:00:01Z

View / edit / reply to this conversation on ReviewNB

OriolAbril commented on 2022-02-27T22:00:01Z
----------------------------------------------------------------

use constant/mutable things to get rid of warning.

review-notebook-app · 2022-02-27T22:00:02Z

View / edit / reply to this conversation on ReviewNB

OriolAbril commented on 2022-02-27T22:00:01Z
----------------------------------------------------------------

Line #8.        sd_dist = pm.Exponential.dist(0.5, shape=2)

Can this be dims="param" so it's clear for readers what each position corresponds to? I'm not sure if dims work with .dist

fonnesbeck commented on 2022-02-27T22:32:40Z
----------------------------------------------------------------

Unfortunately it is not compatible.

review-notebook-app · 2022-02-27T22:00:02Z

View / edit / reply to this conversation on ReviewNB

OriolAbril commented on 2022-02-27T22:00:02Z
----------------------------------------------------------------

This model doesn't look converged :/

cfonnesbeck commented on 2022-03-04T01:19:01Z
----------------------------------------------------------------

Increased tuning seems to fix it.

review-notebook-app · 2022-02-27T22:00:03Z

View / edit / reply to this conversation on ReviewNB

OriolAbril commented on 2022-02-27T22:00:03Z
----------------------------------------------------------------

Adding predictions to the predictions group can be simplified to

pm.sample_posterior_predictive(contextual_effect_trace, predictions=True, extend_inferencedata=True);

fonnesbeck commented on 2022-02-27T22:42:11Z
----------------------------------------------------------------

Unfortunately this does not seem to work. A posterior_predictive group does not get added does not get added to the trace and has to be added manually.

OriolAbril commented on 2022-02-27T22:49:42Z
----------------------------------------------------------------

this should add a predictions group, it seems to be working on the example in the docs: https://docs.pymc.io/en/latest/learn/examples/posterior_predictive.html#prediction. Which groups do you have?

cfonnesbeck commented on 2022-03-04T01:40:04Z
----------------------------------------------------------------

OK, figured it out.

review-notebook-app · 2022-02-27T22:00:04Z

View / edit / reply to this conversation on ReviewNB

OriolAbril commented on 2022-02-27T22:00:04Z
----------------------------------------------------------------

you should also add yourself again, make sure the notebook is executed sequentially until the end and add the page footer cell

fonnesbeck · 2022-02-27T22:29:13Z

Sorry, not sure what you mean by decoy coordinates. Do they not have to be in the coords dict to be used by floor_idx?

View entire conversation on ReviewNB

fonnesbeck · 2022-02-27T22:32:41Z

Unfortunately it is not compatible.

View entire conversation on ReviewNB

fonnesbeck · 2022-02-27T22:42:12Z

Unfortunately this does not seem to work. A posterior_predictive group does not get added does not get added to the trace and has to be added manually.

View entire conversation on ReviewNB

OriolAbril · 2022-02-27T22:47:14Z

no, floor_idx is already an array of the right shape, so dims here have only 2 functions (out of 3 possible functions): annotating the dimension, and checking the shape matches for all objects annotated with that dimension. It is not being used to define the shape like it would in pm.Normal("var", 0, 1, dims="obs_id").

With the improved Aesara capabilities and the refactors to shape/size/dims functionality done by Brandon and Michael (I think it was mostly the two of them) when annotating dimensions, we can use dimension names that are not in coords which is very convenient when used in conjunction with pm.Data because the dimension and the variable where it is first annotated are linked so when generating predictions we only need to update the data instead of both data and coordinate values. See for example https://github.com/pymc-devs/pymc/blob/main/pymc/tests/test_shape_handling.py#L336

View entire conversation on ReviewNB

OriolAbril · 2022-02-27T22:49:43Z

this should add a predictions group, it seems to be working on the example in the docs: https://docs.pymc.io/en/latest/learn/examples/posterior_predictive.html#prediction. Which groups do you have?

View entire conversation on ReviewNB

ghost · 2022-03-04T01:19:02Z

Increased tuning seems to fix it.

View entire conversation on ReviewNB

ghost · 2022-03-04T01:40:05Z

OK, figured it out.

View entire conversation on ReviewNB

fonnesbeck · 2022-03-04T01:51:04Z

OK, should be good to go now, unless there are further issues.

review-notebook-app · 2022-03-04T19:43:41Z

View / edit / reply to this conversation on ReviewNB

OriolAbril commented on 2022-03-04T19:43:41Z
----------------------------------------------------------------

can you use the mutable kwarg to get rid of the warning?

review-notebook-app · 2022-03-04T19:43:42Z

View / edit / reply to this conversation on ReviewNB

OriolAbril commented on 2022-03-04T19:43:42Z
----------------------------------------------------------------

I would be nice to update the coordinate values here now that obs_id is no longer an index but maps directly to counties. I looked at the docstrings and it seems that only model.set_data accepts coords to be updated, but pm.set_data doesn't.

fonnesbeck commented on 2022-04-11T02:24:05Z
----------------------------------------------------------------

Is there anything we can do about this now?

OriolAbril commented on 2022-04-11T21:27:45Z
----------------------------------------------------------------

use the model.set_data method two times or speed https://github.com/pymc-devs/pymc/pull/5588 up

cfonnesbeck commented on 2022-04-12T00:01:55Z
----------------------------------------------------------------

Let's punt on this until set_data is updated.

OriolAbril · 2022-03-04T19:46:27Z

/pre-commit run

Trying to see if this fixes the ci job, you'll have to pull before doing the changes.

OriolAbril · 2022-03-04T19:49:42Z

I never can get pre-commit magic comment to work 🤷

fonnesbeck · 2022-04-11T02:24:06Z

Is there anything we can do about this now?

View entire conversation on ReviewNB

OriolAbril · 2022-04-11T21:27:47Z

use the model.set_data method two times or speed https://github.com/pymc-devs/pymc/pull/5588 up

View entire conversation on ReviewNB

ghost · 2022-04-12T00:01:57Z

Let's punt on this until set_data is updated.

View entire conversation on ReviewNB

* Add guide on how to wrap a JAX function in a Aesara Op * Fix typos and reorder last sections * Fix reference paths * Simplify model and be more verbose in Op creation * Address review comments * Address more review comments * Fix more Aesara docs references * Use reference to Aesara index page * Update myst_nbs/case_studies/wrapping_jax_function.myst.md Co-authored-by: Oriol Abril-Pla <oriol.abril.pla@gmail.com>

* create truncated regression example * delete truncated regression example from main branch * create truncated regression example * delete truncated regression example from main branch * create truncated regression example * delete truncated regression example from main branch * initial commit * update link to pull request in Authors section * add tag `classification` * update authorship verbs * plot through xarray, using XrContinuousRV * add x axis labels * add xarray_einstats to watermark, and fix 'classification' as a tag, not a category Co-authored-by: Benjamin T. Vincent <drbenvincent@users.noreply.github.com>

Add *.DS_Store to .gitignore Co-authored-by: Benjamin T. Vincent <drbenvincent@users.noreply.github.com>

* create truncated regression example * delete truncated regression example from main branch * create truncated regression example * delete truncated regression example from main branch * create truncated regression example * delete truncated regression example from main branch * initial commit * obey the law in terms of numpy random number generation + az style * improvements to schematic figure * replace image with non transparent background * fix embedded image Co-authored-by: Oriol Abril-Pla <oriol.abril.pla@gmail.com> * fix embedded image - not sure why this didn't register before * address all review comments - change notebook tag - remove plt.tight_layout() - ax.legend() -> plt.legend() * make code cell visible * remove 2 unused lines * add explanation tag * add technical note re posterior predictive sampling * addressing Luciano's suggested edits * increase clarity of sentence in introduction * add another explanatory markdown cell on the effect posterior * fix typo Co-authored-by: Benjamin T. Vincent <drbenvincent@users.noreply.github.com> Co-authored-by: Oriol Abril-Pla <oriol.abril.pla@gmail.com>

* initial try * undo spacing error * specify location files * update paths * add exclude notebooks * missing notebooks * undo spacing

* updated notebook * run wit latest v2 version * fixed graphviz redering * partial resolution of pr comments * Add guide on how to wrap a JAX function in a Aesara Op (#302) * Add guide on how to wrap a JAX function in a Aesara Op * Fix typos and reorder last sections * Fix reference paths * Simplify model and be more verbose in Op creation * Address review comments * Address more review comments * Fix more Aesara docs references * Use reference to Aesara index page * Update myst_nbs/case_studies/wrapping_jax_function.myst.md Co-authored-by: Oriol Abril-Pla <oriol.abril.pla@gmail.com> * update twitter link (#314) * update Gaussian Mixture Model example with `pm.NormalMixture` (#310) * create truncated regression example * delete truncated regression example from main branch * create truncated regression example * delete truncated regression example from main branch * create truncated regression example * delete truncated regression example from main branch * initial commit * update link to pull request in Authors section * add tag `classification` * update authorship verbs * plot through xarray, using XrContinuousRV * add x axis labels * add xarray_einstats to watermark, and fix 'classification' as a tag, not a category Co-authored-by: Benjamin T. Vincent <drbenvincent@users.noreply.github.com> * add `*.DS_Store` to `.gitignore` (#315) Add *.DS_Store to .gitignore Co-authored-by: Benjamin T. Vincent <drbenvincent@users.noreply.github.com> * updated notebook * run wit latest v2 version * fixed graphviz redering * partial resolution of pr comments * reverted plot labels * restored plt labels * fixed arviz labels * fixed sampling warning * fixed authors Co-authored-by: Ricardo Vieira <28983449+ricardoV94@users.noreply.github.com> Co-authored-by: Oriol Abril-Pla <oriol.abril.pla@gmail.com> Co-authored-by: Abhishek K M <67158080+Sync271@users.noreply.github.com> Co-authored-by: Benjamin T. Vincent <b.t.vincent@dundee.ac.uk> Co-authored-by: Benjamin T. Vincent <drbenvincent@users.noreply.github.com>

* add bibtex reference for Rasmussen & Williams book * update the latest Mauna Loa C02 data

fonnesbeck · 2022-05-07T17:05:16Z

Should be clean now.

OriolAbril

commented on the myst because it was easier, best to make the changes on from jupyter

.pre-commit-config.yaml

myst_nbs/case_studies/multilevel_modeling.myst.md

fonnesbeck · 2022-05-09T01:59:31Z

I'm getting an erroneous jupytext failure:

The two links in this paragraph are external.

OriolAbril · 2022-05-09T16:11:40Z

These should be flagged, it's the check name that is not right, it finds "urls that should be sphinx cross-references"

The one to the arviz homepage should be {doc}`arviz:index` the one to the xarrayforarviz notebook should be {ref}`xarray_for_arviz`

OriolAbril · 2022-05-09T23:09:52Z

examples/case_studies/multilevel_modeling.ipynb

    "# A Primer on Bayesian Methods for Multilevel Modeling\n",
    "\n",
    ":::{post} 27 February, 2022\n",
-    ":tags: hierarchical, pymc.Data, pymc.Deterministic, pymc.Exponential, pymc.LKJCholeskyCov, pymc.Model, pymc.MvNormal, pymc.Normal\n",
+    ":tags: hierarchical, Data, Deterministic, Exponential, LKJCholeskyCov, Model, MvNormal, Normal\n",


Suggested change

":tags: hierarchical, Data, Deterministic, Exponential, LKJCholeskyCov, Model, MvNormal, Normal\n",

":tags: hierarchical model, case study\n",

OriolAbril · 2022-05-09T23:10:05Z

examples/case_studies/multilevel_modeling.ipynb

@@ -4,10 +4,11 @@
   "cell_type": "markdown",
   "metadata": {},
   "source": [
+    "(notebook_name)=\n",


Suggested change

"(notebook_name)=\n",

"(multilevel_modeling)=\n",

myst_nbs/case_studies/multilevel_modeling.myst.md

OriolAbril · 2022-05-09T23:37:34Z

related to #47

hectormz · 2023-11-07T21:57:27Z

@fonnesbeck I know this is an old PR, but I was wondering if you made the model pngs added in this PR via latex or some other method? Thanks!

Updated multilevel modeling to v4, added missing figures locally

cec20e0

Fixes to multilevel modeling notebook

433b0fc

Fixrs to multilevel modeling in response to feedback

b1639bc

pymc-devs deleted a comment Mar 4, 2022

fonnesbeck added the hold_for_v4 label Mar 4, 2022

OriolAbril removed the hold_for_v4 label Mar 20, 2022

twiecki mentioned this pull request Apr 1, 2022

Have pooling model images be local as they keep disappearing. #282

Closed

fonnesbeck added 2 commits April 10, 2022 21:25

Merge remote-tracking branch 'upstream/main' into multilevel_v4_update

343a12a

Added missing mutable argument to pooled model; re-rendered notebook

8c6451a

OriolAbril mentioned this pull request Apr 11, 2022

Add coords argument to pymc.set_data pymc-devs/pymc#5588

Merged

ricardoV94 and others added 9 commits May 6, 2022 20:11

update twitter link (#314)

2137095

add *.DS_Store to .gitignore (#315)

e4b67c0

Add *.DS_Store to .gitignore Co-authored-by: Benjamin T. Vincent <drbenvincent@users.noreply.github.com>

Add a pre-commit check to prevent urls pointing to pymc docs (#311)

5b854be

* initial try * undo spacing error * specify location files * update paths * add exclude notebooks * missing notebooks * undo spacing

GP Mauna Loa with PyMC3 best practices (#316)

38f2a50

* add bibtex reference for Rasmussen & Williams book * update the latest Mauna Loa C02 data

Pre-commit fixes

20efccc

OriolAbril reviewed May 8, 2022

View reviewed changes

Format fixes; removed multilevel modeling from excluded list

0c322b6

OriolAbril reviewed May 9, 2022

View reviewed changes

OriolAbril added 5 commits May 10, 2022 02:12

Update myst_nbs/case_studies/multilevel_modeling.myst.md

40d1474

Update myst_nbs/case_studies/multilevel_modeling.myst.md

979e517

Merge branch 'main' into multilevel_v4_update

10c9f9f

update notebook

ca33a3c

fix CI

8e352bb

OriolAbril approved these changes May 9, 2022

View reviewed changes

OriolAbril merged commit bea3ac1 into pymc-devs:main May 9, 2022

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Updated multilevel modeling to v4 #285

Updated multilevel modeling to v4 #285

fonnesbeck commented Feb 27, 2022 •

edited

Loading

review-notebook-app bot commented Feb 27, 2022

review-notebook-app bot commented Feb 27, 2022 •

edited

Loading

review-notebook-app bot commented Feb 27, 2022 •

edited

Loading

review-notebook-app bot commented Feb 27, 2022 •

edited

Loading

review-notebook-app bot commented Feb 27, 2022 •

edited

Loading

review-notebook-app bot commented Feb 27, 2022 •

edited

Loading

review-notebook-app bot commented Feb 27, 2022 •

edited

Loading

review-notebook-app bot commented Feb 27, 2022 •

edited

Loading

fonnesbeck commented Feb 27, 2022

fonnesbeck commented Feb 27, 2022

fonnesbeck commented Feb 27, 2022

OriolAbril commented Feb 27, 2022

OriolAbril commented Feb 27, 2022

ghost commented Mar 4, 2022

ghost commented Mar 4, 2022

fonnesbeck commented Mar 4, 2022

review-notebook-app bot commented Mar 4, 2022 •

edited

Loading

review-notebook-app bot commented Mar 4, 2022 •

edited

Loading

OriolAbril commented Mar 4, 2022

OriolAbril commented Mar 4, 2022

fonnesbeck commented Apr 11, 2022

OriolAbril commented Apr 11, 2022

ghost commented Apr 12, 2022

fonnesbeck commented May 7, 2022

OriolAbril left a comment

fonnesbeck commented May 9, 2022

OriolAbril commented May 9, 2022

OriolAbril May 9, 2022

OriolAbril May 9, 2022

OriolAbril commented May 9, 2022

hectormz commented Nov 7, 2023

	":tags: hierarchical, Data, Deterministic, Exponential, LKJCholeskyCov, Model, MvNormal, Normal\n",
	":tags: hierarchical model, case study\n",

Updated multilevel modeling to v4 #285

Updated multilevel modeling to v4 #285

Conversation

fonnesbeck commented Feb 27, 2022 • edited Loading

review-notebook-app bot commented Feb 27, 2022

review-notebook-app bot commented Feb 27, 2022 • edited Loading

review-notebook-app bot commented Feb 27, 2022 • edited Loading

review-notebook-app bot commented Feb 27, 2022 • edited Loading

review-notebook-app bot commented Feb 27, 2022 • edited Loading

review-notebook-app bot commented Feb 27, 2022 • edited Loading

review-notebook-app bot commented Feb 27, 2022 • edited Loading

review-notebook-app bot commented Feb 27, 2022 • edited Loading

fonnesbeck commented Feb 27, 2022

fonnesbeck commented Feb 27, 2022

fonnesbeck commented Feb 27, 2022

OriolAbril commented Feb 27, 2022

OriolAbril commented Feb 27, 2022

ghost commented Mar 4, 2022

ghost commented Mar 4, 2022

fonnesbeck commented Mar 4, 2022

review-notebook-app bot commented Mar 4, 2022 • edited Loading

review-notebook-app bot commented Mar 4, 2022 • edited Loading

OriolAbril commented Mar 4, 2022

OriolAbril commented Mar 4, 2022

fonnesbeck commented Apr 11, 2022

OriolAbril commented Apr 11, 2022

ghost commented Apr 12, 2022

fonnesbeck commented May 7, 2022

OriolAbril left a comment

Choose a reason for hiding this comment

fonnesbeck commented May 9, 2022

OriolAbril commented May 9, 2022

OriolAbril May 9, 2022

Choose a reason for hiding this comment

OriolAbril May 9, 2022

Choose a reason for hiding this comment

OriolAbril commented May 9, 2022

hectormz commented Nov 7, 2023

fonnesbeck commented Feb 27, 2022 •

edited

Loading

review-notebook-app bot commented Feb 27, 2022 •

edited

Loading

review-notebook-app bot commented Feb 27, 2022 •

edited

Loading

review-notebook-app bot commented Feb 27, 2022 •

edited

Loading

review-notebook-app bot commented Feb 27, 2022 •

edited

Loading

review-notebook-app bot commented Feb 27, 2022 •

edited

Loading

review-notebook-app bot commented Feb 27, 2022 •

edited

Loading

review-notebook-app bot commented Feb 27, 2022 •

edited

Loading

review-notebook-app bot commented Mar 4, 2022 •

edited

Loading

review-notebook-app bot commented Mar 4, 2022 •

edited

Loading