Method skew

dask/dataframe/dask_expr/_collection.py:1643–1699

.. note:: This implementation follows the dask.array.stats implementation of skewness and calculates skewness without taking into account a bias term for finite sample size, which corresponds to the default settings of the scipy.stats skewness ca

(
        self,
        axis=0,
        bias=True,
        nan_policy="propagate",
        numeric_only=False,
    )

    @derived_from(pd.DataFrame)
    def skew(
        self,
        axis=0,
        bias=True,
        nan_policy="propagate",
        numeric_only=False,
    ):
        """
        .. note::

           This implementation follows the dask.array.stats implementation
           of skewness and calculates skewness without taking into account
           a bias term for finite sample size, which corresponds to the
           default settings of the scipy.stats skewness calculation. However,
           Pandas corrects for this, so the values differ by a factor of
           (n * (n - 1)) ** 0.5 / (n - 2), where n is the number of samples.

           Further, this method currently does not support filtering out NaN
           values, which is again a difference to Pandas.
        """
        _raise_if_object_series(self, "skew")
        if axis is None:
            raise ValueError("`axis=None` isn't currently supported for `skew`")
        axis = self._validate_axis(axis)

        if is_dataframe_like(self):
            # Let pandas raise errors if necessary
            meta = self._meta_nonempty.skew(axis=axis, numeric_only=numeric_only)
        else:
            meta = self._meta_nonempty.skew()

        if axis == 1:
            return self.map_partitions(
                M.skew,
                meta=meta,
                axis=axis,
                enforce_metadata=False,
            )

        if not bias:
            raise NotImplementedError("bias=False is not implemented.")
        if nan_policy != "propagate":
            raise NotImplementedError(
                "`nan_policy` other than 'propagate' have not been implemented."
            )

        frame = self
        if frame.ndim > 1:
            frame = frame.select_dtypes(
                include=["number", "bool"], exclude=[np.timedelta64]
            )
        m2 = new_collection(Moment(frame, order=2))
        m3 = new_collection(Moment(frame, order=3))
        result = m3 / m2**1.5
        if result.ndim == 1:
            result = result.fillna(0.0)
        return result