Codestin Search App

IntegralIndefinida · 2025-11-16T23:38:41Z

This PR adds a interaction_terms parameter to the het_white function to allow users to choose whether to include interaction terms in White's heteroskedasticity test. This is useful since adding the cross terms consumes degrees of freedom. Additionally, White's heteroscedasticity test with cross terms can also be a specification test, and, according to Richard Harris (as cited in Gujarati's chapter on heteroscedasticity), if we remove the cross terms, it constitutes a pure heteroscedasticity test.

Changes

Added interaction_terms parameter (default True) to statsmodels.stats.diagnostic.het_white
When interaction_terms=False, the test uses only squared terms (x1², x2², ...) without interaction terms (x1x2, x1x3, ...)

Tests

Added test_het_white_no_interaction_terms to verify the interaction_terms=False option
Reference values verified against EViews (I also verified the base variant results)
[ x ] tests added / passed.
[ x ] code/documentation is well formatted.
[ x ] properly formatted commit message. See
NumPy's guide.

Details

Notes:

It is essential that you add a test when making code changes. Tests are not
needed for doc changes.
When adding a new function, test values should usually be verified in another package (e.g., R/SAS/Stata).
When fixing a bug, you must add a test that would produce the bug in main and
then show that it is fixed with the new code.
New code additions must be well formatted. Changes should pass flake8. If on Linux or OSX, you can
verify you changes are well formatted by running
```
git diff upstream/main -u -- "*.py" | flake8 --diff --isolated
```
assuming flake8 is installed. This command is also available on Windows
using the Windows System for Linux once flake8 is installed in the
local Linux environment. While passing this test is not required, it is good practice and it help
improve code quality in statsmodels.
Docstring additions must render correctly, including escapes and LaTeX.

josef-pkt · 2025-11-17T04:25:26Z

statsmodels/stats/tests/test_diagnostic.py

+
+        hw = smdia.het_white(res.resid, res.model.exog,cross_terms=False)
+        hw_values = (
+            13.25091965953952


commas are missing at end of lines

josef-pkt · 2025-11-17T04:25:43Z

statsmodels/stats/diagnostic.py



-def het_white(resid, exog):
+def het_white(resid, exog,cross_terms=True):


space after comma

josef-pkt · 2025-11-17T04:26:56Z

statsmodels/stats/diagnostic.py

    exog : array_like
-        The explanatory variables for the variance. Squares and interaction
-        terms are automatically included in the auxiliary regression.
+        The explanatory variables for the variance. Squares terms are automatically 


keep closer to original just add "by default":

Squares and, by default, interaction terms are automatically included in the auxiliary regression.

josef-pkt · 2025-11-17T04:32:39Z

Looks good overall

I guess this will then be equivalent to
het_breuschpagan(resid, exog**2)

So, not really needed but addition is fine with me.

statsmodels/stats/tests/test_diagnostic.py

IntegralIndefinida · 2025-11-17T19:37:09Z

Thank you for your comments. I also changed the flag to interaction_terms to maintain the same terminology.

josef-pkt · 2025-11-17T20:48:53Z

Thanks,
PR looks good.
Waiting for the CI to finish, but it looks to me it's ready for merging.

josef-pkt · 2025-11-17T21:28:57Z

ci fails

>       hw = smdia.het_white(res.resid, res.model.exog, interaction_terms=False)
             ^^^^^
E       NameError: name 'smdia' is not defined

and one style failure with white space

bashtage

Some changes please.

bashtage · 2025-11-18T09:35:37Z

statsmodels/stats/diagnostic.py

+        i0, i1 = np.triu_indices(nvars0)
+        exog = x[:, i0] * x[:, i1]
+        nobs, nvars = exog.shape
+        assert nvars == nvars0 * (nvars0 - 1) / 2. + nvars0


No asserts please.

It was in the original function, should I remove it anyway?

Yes, please remove.

bashtage · 2025-11-18T09:36:18Z

statsmodels/stats/diagnostic.py

-        terms are automatically included in the auxiliary regression.
+        The explanatory variables for the variance. Squares and, by default,
+        interaction terms are automatically included in the auxiliary regression.
+    interaction_terms : bool, default True


Should probably be False for now since this would change the output of tests without a deprecation.

interaction_term = True
is the current behavior and needed for backwards compatibility and is the "proper" White test

Sorry - read title backward. Yes, True by default.

IntegralIndefinida · 2025-11-18T18:25:32Z

ci fails

>       hw = smdia.het_white(res.resid, res.model.exog, interaction_terms=False)
             ^^^^^
E       NameError: name 'smdia' is not defined

and one style failure with white space

sorry, I'll fix that

bashtage · 2025-11-18T18:29:12Z

Really should add a check that x has a constant, which is required for White's test (even if the original model doesn't).

IntegralIndefinida · 2025-11-18T19:20:28Z

Really should add a check that x has a constant, which is required for White's test (even if the original model doesn't).

That's done by _check_het_test:

def _check_het_test(x: np.ndarray, test_name: str) -> None:
    """
    Check validity of the exogenous regressors in a heteroskedasticity test

    Parameters
    ----------
    x : ndarray
        The exogenous regressor array
    test_name : str
        The test name for the exception
    """
    x_max = x.max(axis=0)
    if (
        not np.any(((x_max - x.min(axis=0)) == 0) & (x_max != 0))
        or x.shape[1] < 2
    ):
        raise ValueError(
            f"{test_name} test requires exog to have at least "
            "two columns where one is a constant."
        )

IntegralIndefinida · 2025-11-18T19:27:55Z

I also removed the line

question: does f-statistic make sense? constant ?

because, as @bashtage says, a constant term is required and the F-statistic indeed makes sense, since the White's test only distributes $\chi^2$ asymptotically.

bashtage

One small change to improve doc generation.

bashtage · 2025-11-26T16:59:31Z

statsmodels/stats/diagnostic.py

-
    References
    ----------
    Greene section 11.4.1 5th edition p. 222. Test statistic reproduces


Could we change the references to have the propper format. Should be like

.. [1] Greene, William H. Econometric analysis. 5th Edition. Pearson Education, 2002.
.. [2] Damodar N. Gujarati, Basic Econometrics, section 11.5. Pg 387.

Sure, that'd be better

bashtage · 2025-11-26T17:02:30Z

Close and reopen to force CI run

… into het_white

bashtage · 2025-11-26T19:23:57Z

statsmodels/stats/diagnostic.py:827:1: E302 expected 2 blank lines, found 1

Lint failure

IntegralIndefinida · 2025-11-30T22:58:53Z

I fixed the missing line, but the test still fails with the following linting errors:

Running flake8 linting
Linting all files with limited rules
statsmodels/discrete/discrete_model.py:522:13: B043 Do not call delattr with a constant attribute value, it is not any safer than normal property access.
statsmodels/discrete/discrete_model.py:1054:13: B043 Do not call delattr with a constant attribute value, it is not any safer than normal property access.
statsmodels/discrete/discrete_model.py:1056:13: B043 Do not call delattr with a constant attribute value, it is not any safer than normal property access.
statsmodels/genmod/generalized_linear_model.py:375:13: B043 Do not call delattr with a constant attribute value, it is not any safer than normal property access.
statsmodels/genmod/generalized_linear_model.py:377:13: B043 Do not call delattr with a constant attribute value, it is not any safer than normal property access.
Changed files failed linting using the required set of rules.
Additions and changes must conform to Python code style rules.
No new files to lint
Running isort
Skipped 1 files

##[error]Bash exited with code '1'.
Finishing: Check style

Those are unrelated to my commits

bashtage · 2026-01-08T13:08:41Z

Closing and reopening to see CI run

ENH: Add no cross terms option to White's test for heteroscedasticity

309867b

josef-pkt reviewed Nov 17, 2025

View reviewed changes

statsmodels/stats/diagnostic.py Outdated

def het_white(resid, exog):

def het_white(resid, exog,cross_terms=True):

Copy link

Member

josef-pkt Nov 17, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

space after comma

josef-pkt reviewed Nov 17, 2025

View reviewed changes

github-advanced-security bot found potential problems Nov 17, 2025

View reviewed changes

statsmodels/stats/tests/test_diagnostic.py Fixed Show fixed Hide fixed

IntegralIndefinida force-pushed the het_white branch from c9df30d to 9855622 Compare November 17, 2025 19:31

Address review: fixes typos and style issues

43be2d3

IntegralIndefinida force-pushed the het_white branch from 9855622 to 43be2d3 Compare November 17, 2025 19:35

bashtage requested changes Nov 18, 2025

View reviewed changes

Address review 2: fixes more code errors and removes asserts

19d9eba

IntegralIndefinida requested a review from bashtage November 18, 2025 19:38

Removes trailing whitespace

8d16ba5

bashtage requested changes Nov 26, 2025

View reviewed changes

bashtage closed this Nov 26, 2025

bashtage reopened this Nov 26, 2025

IntegralIndefinida added 2 commits November 26, 2025 14:05

Address review 3: Changes References formatting

89c70c7

Merge branch 'het_white' of github.com:IntegralIndefinida/statsmodels…

3b86cf9

… into het_white

IntegralIndefinida closed this Nov 26, 2025

IntegralIndefinida reopened this Nov 26, 2025

Fixes lint Failure: Adds missing blank line

51249c8

IntegralIndefinida closed this Nov 26, 2025

IntegralIndefinida reopened this Nov 26, 2025

IntegralIndefinida closed this Nov 30, 2025

IntegralIndefinida reopened this Nov 30, 2025

bashtage closed this Jan 8, 2026

bashtage reopened this Jan 8, 2026



		def het_white(resid, exog):
		def het_white(resid, exog,cross_terms=True):

Conversation

IntegralIndefinida commented Nov 16, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

josef-pkt commented Nov 17, 2025

Uh oh!

Uh oh!

IntegralIndefinida commented Nov 17, 2025

Uh oh!

josef-pkt commented Nov 17, 2025

Uh oh!

josef-pkt commented Nov 17, 2025

Uh oh!

bashtage left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

IntegralIndefinida commented Nov 18, 2025

Uh oh!

bashtage commented Nov 18, 2025

Uh oh!

IntegralIndefinida commented Nov 18, 2025

Uh oh!

IntegralIndefinida commented Nov 18, 2025

Uh oh!

bashtage left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

bashtage commented Nov 26, 2025

Uh oh!

bashtage commented Nov 26, 2025

Uh oh!

IntegralIndefinida commented Nov 30, 2025

Uh oh!

bashtage commented Jan 8, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

IntegralIndefinida commented Nov 16, 2025 •

edited

Loading