Skip to content

Adds triangular option to DataFrame.corr, closes #22840 #22842

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
wants to merge 9 commits into from
Closed

Adds triangular option to DataFrame.corr, closes #22840 #22842

Show file tree
Hide file tree
Changes from 4 commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
8a2eac3
modify DataFrame.corr to allow lower triangular correlation matrix
Sep 26, 2018
9a95e8c
add test for DataFrame.corr(tri=True)
Sep 26, 2018
9acc618
PR #22840
Sep 26, 2018
c97fec1
PR #22840
Sep 26, 2018
9b0276a
PR #22842
Sep 26, 2018
b15436b
PR #22840
Sep 26, 2018
6be67f6
modify DataFrame.corr to allow triangular correlation matrix
Sep 26, 2018
d619d50
PR #22842: rebase
Sep 27, 2018
9df2bcb
PR #22842: add to DataFrame.corr docstring
Sep 28, 2018
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions doc/source/whatsnew/v0.24.0.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,7 @@ New features
the user to override the engine's default behavior to include or omit the
dataframe's indexes from the resulting Parquet file. (:issue:`20768`)
- :meth:`DataFrame.corr` and :meth:`Series.corr` now accept a callable for generic calculation methods of correlation, e.g. histogram intersection (:issue:`22684`)
- :meth:`DataFrame.corr` now accepts an argument indicating whether or not a triangular correlation matrix should be returned (:issue:`22840`)


.. _whatsnew_0240.enhancements.extension_array_operators:
Expand Down
26 changes: 24 additions & 2 deletions pandas/core/frame.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6667,7 +6667,7 @@ def _series_round(s, decimals):
# ----------------------------------------------------------------------
# Statistical methods, etc.

def corr(self, method='pearson', min_periods=1):
def corr(self, method='pearson', min_periods=1, tri=None):
"""
Compute pairwise correlation of columns, excluding NA/null values

Expand All @@ -6686,6 +6686,12 @@ def corr(self, method='pearson', min_periods=1):
to have a valid result. Currently only available for pearson
and spearman correlation

tri : {'upper', 'lower'} or None, default : None
Whether or not to return the upper / lower triangular
correlation matrix
Copy link
Contributor

Choose a reason for hiding this comment

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

Could you add an explanation for why a user might want to do this? And an example below using upper or lower?

Copy link
Member Author

Choose a reason for hiding this comment

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

Sure, you mean within the docstring itself right?


.. versionadded:: 0.24.0

Copy link
Member

Choose a reason for hiding this comment

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

Can you add a versionadded directive here?

Returns
-------
y : DataFrame
Expand Down Expand Up @@ -6741,7 +6747,23 @@ def corr(self, method='pearson', min_periods=1):
"'spearman', or 'kendall', '{method}' "
"was supplied".format(method=method))

return self._constructor(correl, index=idx, columns=cols)
corr_mat = self._constructor(correl, index=idx, columns=cols)

if tri is not None:
mask = np.tril(np.ones_like(corr_mat,
Copy link
Member

Choose a reason for hiding this comment

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

Is it possible to use correl instead of corr_mat here to avoid the conversion hit of DataFrame to numpy array when np.ones_like is called? (I think correl should be a numpy array here)

Copy link
Member Author

Choose a reason for hiding this comment

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

Yes, nice catch. correl should already be a numpy array.

dtype=np.bool),
k=-1)

if tri == 'lower':
return corr_mat.where(mask)
Copy link
Contributor

Choose a reason for hiding this comment

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

Not the speed is probably important, but corr+np.tril(corr*np.nan) is probably faster than indexing.

elif tri == 'upper':
return corr_mat.where(mask.T)
else:
raise ValueError("tri must be either 'lower', "
"or 'upper', '{tri_method}' "
"was supplied".format(tri_method=tri))
else:
return corr_mat

def cov(self, min_periods=None):
"""
Expand Down
12 changes: 12 additions & 0 deletions pandas/tests/frame/test_analytics.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -138,6 +138,18 @@ def test_corr_invalid_method(self):
with tm.assert_raises_regex(ValueError, msg):
df.corr(method="____")

def test_corr_tril(self):
Copy link
Member

Choose a reason for hiding this comment

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

For these tests, it's better to build an expected DataFrame output and compare it to a resulting DataFrame. You can build a DataFrame that gives a trivial correlation matrix (like a np.ones DataFrame)

result = df.corr(tri=...)
expected = pd.DataFrame(...)
tm.assert_frame_equal(result, expected)

Copy link
Member Author

@dsaxton dsaxton Sep 26, 2018 •
edited
Loading

Choose a reason for hiding this comment

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

I updated the tests according to this idea. We're still using DataFrames inside ones_like, but I figured this is okay since they're small test cases.

# GH PR #22840
df = pd.DataFrame(np.random.normal(size=(100, 5)))
corr_mat = df.corr(tri='lower')
assert corr_mat.notnull().sum().sum() == 10

def test_corr_triu(self):
# GH PR #22840
df = pd.DataFrame(np.random.normal(size=(100, 5)))
corr_mat = df.corr(tri='upper')
assert corr_mat.notnull().sum().sum() == 10

def test_cov(self):
# min_periods no NAs (corner case)
expected = self.frame.cov()
Expand Down