Add Notebook on `Stargazer` and `pymarginaleffects` support to the docs #473

s3alfisc · 2024-05-31T14:30:44Z

Add jupyter notebooks that shows how to use Stargazer and marginaleffects with pyfixest.

juanitorduz · 2024-06-07T21:17:03Z

Cool PR!

The error from the CI looks weird.

____________________ ERROR collecting tests/test_confint.py ____________________
ImportError while importing test module '/home/runner/work/pyfixest/pyfixest/tests/test_confint.py'.
Hint: make sure your test modules/packages have valid Python names.
Traceback:
/opt/hostedtoolcache/Python/3.12.3/x64/lib/python3.12/importlib/__init__.py:90: in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
tests/test_confint.py:1: in <module>
    import doubleml as dml
../../../.cache/pypoetry/virtualenvs/pyfixest-va_uSH9_-py3.12/lib/python3.12/site-packages/doubleml/__init__.py:1: in <module>
    from pkg_resources import get_distribution
E   ModuleNotFoundError: No module named 'pkg_resources'

I have seen DoubleML does not support python 3.12 https://github.com/DoubleML/doubleml-for-py/blob/main/setup.py

Maybe see https://stackoverflow.com/questions/7446187/no-module-named-pkg-resources

s3alfisc · 2024-06-08T10:59:44Z

Thanks for taking a look at the error! DoubelML is only used for testing simultaneous confidence intervals - I think I should just rework the tests to only run the DoubleML implementation once and then save it to a file that the tests can read from. This way, there's less room for error!

Btw, I am very excited about Stargazer support, I think that it produces very pretty tables =)

s3alfisc · 2024-06-09T18:01:19Z

This PR adds an internal Stargazer class that inherits from Stargazer, but customizes the regression output for the PyFixest use case by adding information on the employed fixed effects:

import pyfixest as pf

fit = pf.feols("Y ~ X1 | csw(f1, f2)"
Stargazer(fit.to_list())

will return a Stargazer table including info on the employed fixed effects.

@aeturrell, are there any other utility functions you'd like to see? E.g. we could add methods to easily add randomization inference / bootstrap p-values, but given that Stargazer.add_lines() is so convenient to use, I wonder if it is already overkill?

s3alfisc · 2024-06-09T18:09:05Z

Juan (@juanitorduz ) - is it considered bad practice to exclude dev dependencies from the pyproject.toml? If no, I would just drop doubleml from the dependencies. For testing against doubleml, I am now saving a static datafile in tests/data/dml_res.csv, and in principle, doubleml no longer has to be available when running the tests. But excluding it would mean that effectively, users couldn't replicate the entire test suite end to end in the default poetry environment of pyfixest.

juanitorduz · 2024-06-09T18:14:44Z

Juan (@juanitorduz ) - is it considered bad practice to exclude dev dependencies from the pyproject.toml? If no, I would just drop doubleml from the dependencies. For testing against doubleml, I am now saving a static datafile in tests/data/dml_res.csv, and in principle, doubleml no longer has to be available when running the tests. But excluding it would mean that effectively, users couldn't replicate the entire test suite end to end in the default poetry environment of pyfixest.

This sounds like a hack 😄. What about just keeping it in and skipping the tests for the specific Python version using skipif as described in https://docs.pytest.org/en/stable/how-to/skipping.html#id1 ?

juanitorduz · 2024-06-09T20:29:32Z

tests/test_confint.py

@@ -47,6 +50,7 @@ def test_confint():
    assert np.all(confint1 != confint3)


+@pytest.mark.skipif(sys.version_info >= (3, 12), reason="requires python3.11 or lower.")


What if you import doubleml inside the test test_against_doubleml so that it does not get imported globally? I think the error is even at import time

It worked 🎉!

Yes! Was just about to comment 😀 thanks for your help! Now I need to work towards acceptance of the stargazer pr. Do you have any feedback on the content of the new notebooks? Both of them might need some polish here and there.

I manly looked into the CI issue. I can take a look into the notebooks themselves either today or tomorrow :)

juanitorduz · 2024-06-10T07:56:12Z

@s3alfisc did you remove the ReviewNB app?

aeturrell · 2024-06-10T08:54:55Z

This PR adds an internal Stargazer class that inherits from Stargazer, but customizes the regression output for the PyFixest use case by adding information on the employed fixed effects:
import pyfixest as pf

fit = pf.feols("Y ~ X1 | csw(f1, f2)"
Stargazer(fit.to_list())
will return a Stargazer table including info on the employed fixed effects.

@aeturrell, are there any other utility functions you'd like to see? E.g. we could add methods to easily add randomization inference / bootstrap p-values, but given that Stargazer.add_lines() is so convenient to use, I wonder if it is already overkill?

Oh wow, this is so cool! The only other metrics I can think of that you sometimes see on these tables is the BIC and the pseudo or adjusted R2. Though the latter sometimes replaces R2. But agree with your instinct that if it's easy to add extra lines then it's not something I'd worry too much about.

Once this is merged in and released, let me know, and I will make sure I (eventually!) update Coding for Economists to show how stargazer and pymarginaleffects can be used with pyfixest.

s3alfisc · 2024-06-10T10:09:13Z

@s3alfisc did you remove the ReviewNB app?

I was wondering about this as well - I think it didn't survive the migration to py-econometrics. Will fix this after work!

s3alfisc · 2024-06-10T18:26:01Z

I was wondering about this as well - I think it didn't survive the migration to py-econometrics. Will fix this after work!

Yes, that was it - you can take a look at the notebooks here.

The only other metrics I can think of that you sometimes see on these tables is the BIC and the pseudo or adjusted R2.

None of these are implemented yet, unfortunately 😅 I was thinking that it could be a nice add-on if pf.Stargazer allowed to easily add custom statistics in the right place - i.e. if randomization inference has been computed, to have a method add_line_ritest() that then adds the RI p-values in the right place. But not something for this PR I think =)

juanitorduz · 2024-06-11T15:10:17Z

@s3alfisc This week I am swamped with work 😢 . Fel free to merge this one as it overall looks good! I can take a better look next week and open a PR if necessary :) (when ReviewNB is back XD )

s3alfisc · 2024-06-11T21:19:15Z

This week I am swamped with work

No worries! I agree that both should be fine =)

Btw, the notebooks are back, it was indeed the package migration.

docs/marginaleffects.ipynb

juanitorduz · 2024-07-01T19:50:07Z

docs/marginaleffects.ipynb

@@ -0,0 +1,371 @@
+{


For the non-expert readers, I suggest you explain how you reach this conclusion from the table :) For example, is it the -0.81?

Reply via ReviewNB

docs/marginaleffects.ipynb

juanitorduz · 2024-07-01T19:52:05Z

@s3alfisc Apologies for the late review! I did not know about this integration with the py-marginal effects package! Super useful! I just added some suggestions to make the content more welcoming, especially for new users who are not familiar with the topic 💪

Other than that I think we should share this to the (Python) world :D

s3alfisc · 2024-07-01T20:48:42Z

@s3alfisc Apologies for the late review! I did not know about this integration with the py-marginal effects package! Super useful! I just added some suggestions to make the content more welcoming, especially for new users who are not familiar with the topic 💪

Thanks Juan! No worries at all - last week, it was me who was really swamped with work, so I can very much relate! I'll tackle your comments throughout the week =)

Other than that I think we should share this to the (Python) world :D

Yes, we should! I thing that py-marginaleffects is completely flying under the radar for some reason. I guess that after I polish the marginaleffects notebook, I could use it to promote the package a little bit =)

juanitorduz · 2024-07-02T12:48:02Z

Yes! PLease do so! I want to explore the package integration further and most likely expand the examples :)

aeturrell · 2024-07-03T09:16:07Z

When this is merged and out, will give it a boost so people are aware. Very cool.

review-notebook-app · 2024-07-07T13:44:38Z

Check out this pull request on

See visual diffs & provide feedback on Jupyter Notebooks.

Powered by ReviewNB

juanitorduz · 2024-07-08T08:00:57Z

@s3alfisc I think the notebooks look good! I would suggest we take an iterative approach and merge -> collect feedback -> iterate :) I think many people would just benefit by knowing we can integrate pyfixest and marginal effects

s3alfisc · 2024-07-08T17:26:28Z

Thanks for your feedback Juan! I agree that an iterative approach is the most reasonable one here =) will merge now!

s3alfisc added 4 commits May 31, 2024 16:18

update docs

c158189

add stargazer to dev dependencies

e97903f

update stargazer vignette

dd386e5

write vignettes for Stargazer and pymarginaleffects

876d6ed

s3alfisc changed the title ~~Add Section on Stargazer support to the docs~~ Add Notebook on Stargazer and pymarginaleffects support to the docs Jun 3, 2024

s3alfisc linked an issue Jun 3, 2024 that may be closed by this pull request

Docs: Write Vignette for pymarginaleffects #274

Closed

s3alfisc added 5 commits June 8, 2024 13:42

move tests against doubleml to static file

2b54c52

add stargazer class

c2fd7c7

get internal Stargazer class

a82430f

resolve merge conflicts

fda2f4c

only return fixef into if at least one model has fixed effects

e908679

s3alfisc added 2 commits June 9, 2024 22:09

skip tests against doublme for py > 3.11

389cb46

delete r data set for doubleml tests

2077a6b

juanitorduz reviewed Jun 9, 2024

View reviewed changes

load doubleml only within test

4d85bed

juanitorduz reviewed Jul 1, 2024

View reviewed changes

minor update on stargazer notebook

f894f6c

add section on Delta method

ece9fc1

s3alfisc added 2 commits July 8, 2024 19:30

solve merge conflicts

456e65d

update lock file

a254f1c

s3alfisc merged commit 956c93e into master Jul 8, 2024
6 checks passed

s3alfisc deleted the stargazer-support branch July 8, 2024 19:22

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add Notebook on `Stargazer` and `pymarginaleffects` support to the docs #473

Add Notebook on `Stargazer` and `pymarginaleffects` support to the docs #473

s3alfisc commented May 31, 2024 •

edited

Loading

juanitorduz commented Jun 7, 2024

s3alfisc commented Jun 8, 2024 •

edited

Loading

s3alfisc commented Jun 9, 2024 •

edited

Loading

s3alfisc commented Jun 9, 2024

juanitorduz commented Jun 9, 2024

juanitorduz Jun 9, 2024

juanitorduz Jun 10, 2024

s3alfisc Jun 10, 2024

juanitorduz Jun 10, 2024

juanitorduz commented Jun 10, 2024

aeturrell commented Jun 10, 2024

s3alfisc commented Jun 10, 2024

s3alfisc commented Jun 10, 2024 •

edited

Loading

juanitorduz commented Jun 11, 2024

s3alfisc commented Jun 11, 2024

juanitorduz Jul 1, 2024 •

edited

Loading

juanitorduz commented Jul 1, 2024 •

edited

Loading

s3alfisc commented Jul 1, 2024 •

edited

Loading

juanitorduz commented Jul 2, 2024

aeturrell commented Jul 3, 2024

review-notebook-app bot commented Jul 7, 2024

juanitorduz commented Jul 8, 2024

s3alfisc commented Jul 8, 2024

		@@ -47,6 +50,7 @@ def test_confint():
		assert np.all(confint1 != confint3)


		@pytest.mark.skipif(sys.version_info >= (3, 12), reason="requires python3.11 or lower.")

Add Notebook on Stargazer and pymarginaleffects support to the docs #473

Add Notebook on Stargazer and pymarginaleffects support to the docs #473

Conversation

s3alfisc commented May 31, 2024 • edited Loading

juanitorduz commented Jun 7, 2024

s3alfisc commented Jun 8, 2024 • edited Loading

s3alfisc commented Jun 9, 2024 • edited Loading

s3alfisc commented Jun 9, 2024

juanitorduz commented Jun 9, 2024

juanitorduz Jun 9, 2024

Choose a reason for hiding this comment

juanitorduz Jun 10, 2024

Choose a reason for hiding this comment

s3alfisc Jun 10, 2024

Choose a reason for hiding this comment

juanitorduz Jun 10, 2024

Choose a reason for hiding this comment

juanitorduz commented Jun 10, 2024

aeturrell commented Jun 10, 2024

s3alfisc commented Jun 10, 2024

s3alfisc commented Jun 10, 2024 • edited Loading

juanitorduz commented Jun 11, 2024

s3alfisc commented Jun 11, 2024

juanitorduz Jul 1, 2024 • edited Loading

Choose a reason for hiding this comment

juanitorduz commented Jul 1, 2024 • edited Loading

s3alfisc commented Jul 1, 2024 • edited Loading

juanitorduz commented Jul 2, 2024

aeturrell commented Jul 3, 2024

review-notebook-app bot commented Jul 7, 2024

juanitorduz commented Jul 8, 2024

s3alfisc commented Jul 8, 2024

Add Notebook on `Stargazer` and `pymarginaleffects` support to the docs #473

Add Notebook on `Stargazer` and `pymarginaleffects` support to the docs #473

s3alfisc commented May 31, 2024 •

edited

Loading

s3alfisc commented Jun 8, 2024 •

edited

Loading

s3alfisc commented Jun 9, 2024 •

edited

Loading

s3alfisc commented Jun 10, 2024 •

edited

Loading

juanitorduz Jul 1, 2024 •

edited

Loading

juanitorduz commented Jul 1, 2024 •

edited

Loading

s3alfisc commented Jul 1, 2024 •

edited

Loading