Fixing gh-17749 requires a small API change to `stats.moment`
Hey all, I suggested a rather simple fix for [gh-17749][1], and it's a sort of non-breaking API change. I was replied in a comment that such a change should go through this mailing list. Here's the Gist of the fix, without unit tests and without docs updates. I will appreciate comments/reviews: [1]: https://github.com/scipy/scipy/issues/17749 ```patch commit f16d2282bd34726960d511c1ed277ba2754d285e Author: Doron Behar <doron.behar@gmail.com> Date: Sat Feb 4 14:50:02 2023 +0200 stats: Allow computing non central moments diff --git a/result b/result new file mode 120000 index 000000000..5cff41b90 --- /dev/null +++ b/result @@ -0,0 +1 @@ +/nix/store/38g1b2ai2k35kq7jw0ffm5la90zaqa0h-python3.10-scipy-1.11.0-dirty-flake \ No newline at end of file diff --git a/scipy/stats/_mstats_basic.py b/scipy/stats/_mstats_basic.py index 2d10561d2..7676c7798 100644 --- a/scipy/stats/_mstats_basic.py +++ b/scipy/stats/_mstats_basic.py @@ -2551,7 +2551,7 @@ def winsorize(a, limits=None, inclusive=(True, True), inplace=False, upinc, contains_nan, nan_policy) -def moment(a, moment=1, axis=0): +def moment(a, moment=1, mean=None, axis=0): """ Calculates the nth moment about the mean for a sample. @@ -2561,13 +2561,15 @@ def moment(a, moment=1, axis=0): data moment : int, optional order of central moment that is returned + mean : float, optional + If True, then the moments are computed centralized. Default is True. axis : int or None, optional Axis along which the central moment is computed. Default is 0. If None, compute over the whole array `a`. Returns ------- - n-th central moment : ndarray or float + n-th (central) moment : ndarray or float The appropriate moment along the given axis or over all values if axis is None. The denominator for the moment calculation is the number of observations, no degrees of freedom correction is done. @@ -2592,8 +2594,11 @@ def moment(a, moment=1, axis=0): # for array_like moment input, return a value for each. if not np.isscalar(moment): - mean = a.mean(axis, keepdims=True) - mmnt = [_moment(a, i, axis, mean=mean) for i in moment] + if mean is None: + _mean = a.mean(axis, keepdims=True) + else: + _mean = mean + mmnt = [_moment(a, i, axis, mean=_mean) for i in moment] return ma.array(mmnt) else: return _moment(a, moment, axis) diff --git a/scipy/stats/_stats_py.py b/scipy/stats/_stats_py.py index 32445bedf..70d098baa 100644 --- a/scipy/stats/_stats_py.py +++ b/scipy/stats/_stats_py.py @@ -1046,14 +1046,14 @@ def moment(a, moment=1, axis=0, nan_policy='propagate'): Notes ----- - The k-th central moment of a data sample is: + The k-th moment of a data sample is: .. math:: - m_k = \frac{1}{n} \sum_{i = 1}^n (x_i - \bar{x})^k + m_k = \frac{1}{n} \sum_{i = 1}^n (x_i)^k - Where n is the number of samples and x-bar is the mean. This function uses - exponentiation by squares [1]_ for efficiency. + Where n is the number of samples. This function uses exponentiation by + squares [1]_ for efficiency. Note that, if `a` is an empty array (``a.size == 0``), array `moment` with one element (`moment.size == 1`) is treated the same as scalar `moment` ```
Hi team, To summarize, the issue (https://github.com/scipy/scipy/issues/17749) requested that `stats.moment` be enhanced to allow the computation of non-central sample moments. The proposed solution is to add an optional keyword argument `center` (or `mean` in the original email). - For central moments (current behavior), use the default value (`None`). The sample mean will be calculated from the provided sample (as it is now). - For efficiency, pass the sample mean (if it is already available). - For non-central moments, pass `0`. Since this requires an addition to the API, we'd like to know whether the proposed enhancement sounds good and, if so, what the new parameter should be called. Matt On Sat, Mar 11, 2023 at 1:54 PM Doron Behar <doron.behar@gmail.com> wrote:
Hey all,
I suggested a rather simple fix for [gh-17749][1], and it's a sort of non-breaking API change. I was replied in a comment that such a change should go through this mailing list. Here's the Gist of the fix, without unit tests and without docs updates. I will appreciate comments/reviews:
[1]: https://github.com/scipy/scipy/issues/17749
```patch commit f16d2282bd34726960d511c1ed277ba2754d285e Author: Doron Behar <doron.behar@gmail.com> Date: Sat Feb 4 14:50:02 2023 +0200
stats: Allow computing non central moments
diff --git a/result b/result new file mode 120000 index 000000000..5cff41b90 --- /dev/null +++ b/result @@ -0,0 +1 @@
+/nix/store/38g1b2ai2k35kq7jw0ffm5la90zaqa0h-python3.10-scipy-1.11.0-dirty-flake \ No newline at end of file diff --git a/scipy/stats/_mstats_basic.py b/scipy/stats/_mstats_basic.py index 2d10561d2..7676c7798 100644 --- a/scipy/stats/_mstats_basic.py +++ b/scipy/stats/_mstats_basic.py @@ -2551,7 +2551,7 @@ def winsorize(a, limits=None, inclusive=(True, True), inplace=False, upinc, contains_nan, nan_policy)
-def moment(a, moment=1, axis=0): +def moment(a, moment=1, mean=None, axis=0): """ Calculates the nth moment about the mean for a sample.
@@ -2561,13 +2561,15 @@ def moment(a, moment=1, axis=0): data moment : int, optional order of central moment that is returned + mean : float, optional + If True, then the moments are computed centralized. Default is True. axis : int or None, optional Axis along which the central moment is computed. Default is 0. If None, compute over the whole array `a`.
Returns ------- - n-th central moment : ndarray or float + n-th (central) moment : ndarray or float The appropriate moment along the given axis or over all values if axis is None. The denominator for the moment calculation is the number of observations, no degrees of freedom correction is done. @@ -2592,8 +2594,11 @@ def moment(a, moment=1, axis=0):
# for array_like moment input, return a value for each. if not np.isscalar(moment): - mean = a.mean(axis, keepdims=True) - mmnt = [_moment(a, i, axis, mean=mean) for i in moment] + if mean is None: + _mean = a.mean(axis, keepdims=True) + else: + _mean = mean + mmnt = [_moment(a, i, axis, mean=_mean) for i in moment] return ma.array(mmnt) else: return _moment(a, moment, axis) diff --git a/scipy/stats/_stats_py.py b/scipy/stats/_stats_py.py index 32445bedf..70d098baa 100644 --- a/scipy/stats/_stats_py.py +++ b/scipy/stats/_stats_py.py @@ -1046,14 +1046,14 @@ def moment(a, moment=1, axis=0, nan_policy='propagate'):
Notes ----- - The k-th central moment of a data sample is: + The k-th moment of a data sample is:
.. math::
- m_k = \frac{1}{n} \sum_{i = 1}^n (x_i - \bar{x})^k + m_k = \frac{1}{n} \sum_{i = 1}^n (x_i)^k
- Where n is the number of samples and x-bar is the mean. This function uses - exponentiation by squares [1]_ for efficiency. + Where n is the number of samples. This function uses exponentiation by + squares [1]_ for efficiency.
Note that, if `a` is an empty array (``a.size == 0``), array `moment` with one element (`moment.size == 1`) is treated the same as scalar `moment` ``` _______________________________________________ SciPy-Dev mailing list -- scipy-dev@python.org To unsubscribe send an email to scipy-dev-leave@python.org https://mail.python.org/mailman3/lists/scipy-dev.python.org/ Member address: mhaberla@calpoly.edu
-- Matt Haberland Assistant Professor BioResource and Agricultural Engineering 08A-3K, Cal Poly
participants (2)
-
Doron Behar -
Matt Haberland