et a standardized
      or vice versa
    - ``'quadrature'``: numerically integrate according to the definition

    Not all `method` options are available for all orders, kinds, and
    distributions. If the selected `method` is not available, a
    ``NotImplementedError`` will be raised.

Returns
-------
out : array
    The moment of the random variable of the specified order and kind.

See Also
--------
pdf
mean
variance
standard_deviation
skewness
kurtosis

Notes
-----
Not all distributions have finite moments of all orders; moments of some
orders may be undefined or infinite. If a formula for the moment is not
specifically implemented for the chosen distribution, SciPy will attempt
to compute the moment via a generic method, which may yield a finite
result where none exists. This is not a critical bug, but an opportunity
for an enhancement.

The definition of a raw moment in the summary is specific to the raw moment
about the origin. The raw moment about any point :math:`a` is:

.. math::

    E[(X-a)^n] = \int_{\chi} (x-a)^n f(x) dx

In this notation, a raw moment about the origin is :math:`\mu'_n = E[x^n]`,
and a central moment is :math:`\mu_n = E[(x-\mu)^n]`, where :math:`\mu`
is the first raw moment; i.e. the mean.

The ``'transform'`` method takes advantage of the following relationships
between moments taken about different points :math:`a` and :math:`b`.

.. math::

    E[(X-b)^n] =  \sum_{i=0}^n E[(X-a)^i] {n \choose i} (a - b)^{n-i}

For instance, to transform the raw moment to the central moment, we let
:math:`b = \mu` and :math:`a = 0`.

The distribution infrastructure provides flexibility for distribution
authors to implement separate formulas for raw moments, central moments,
and standardized moments of any order. By default, the moment of the
desired order and kind is evaluated from the formula if such a formula
is available; if not, the infrastructure uses any formulas that are
available rather than resorting directly to numerical integration.
For instance, if formulas for the first three raw moments are
available and the third standardized moments is desired, the
infrastructure will evaluate the raw moments and perform the transforms
and standardization required. The decision tree is somewhat complex,
but the strategy for obtaining a moment of a given order and kind
(possibly as an intermediate step due to the recursive nature of the
transform formula above) roughly follows this order of priority:

#. Use cache (if order of same moment and kind has been calculated)
#. Use formula (if available)
#. Transform between raw and central moment and/or normalize to convert
   between central and standardized moments (if efficient)
#. Use a generic result true for most distributions (if available)
#. Use quadrature

References
----------
.. [1] Moment, *Wikipedia*,
       https://en.wikipedia.org/wiki/Moment_(mathematics)

Examples
--------
Instantiate a distribution with the desired parameters:

>>> from scipy import stats
>>> X = stats.Normal(mu=1., sigma=2.)

Evaluate the first raw moment:

>>> X.moment(order=1, kind='raw')
1.0
>>> X.moment(order=1, kind='raw') == X.mean() == X.mu
True

Evaluate the second central moment:

>>> X.moment(order=2, kind='central')
4.0
>>> X.moment(order=2, kind='central') == X.variance() == X.sigma**2
True

Evaluate the fourth standardized moment:

>>> X.moment(order=4, kind='standardized')
3.0
>>> X.moment(order=4, kind='standardized') == X.kurtosis(convention='non-excess')
True

r