Getting started


For a quick and hands-on introduction, start with the multinorm.ipynb Jupyter notebook tutorial, then continue reading here.


The multinorm package offers a single class MultiNorm, so you always start like this:

from multinorm import MultiNorm


To create a MultiNorm object, pass a mean vector, a covariance matrix (both as Numpy arrays) and optionally a list of parameter names:

from multinorm import MultiNorm
mean = [10, 20, 30]
covariance = [[1, 0, 0], [0, 4, 0], [0, 0, 9]]
names = ["a", "b", "c"]
mn = MultiNorm(mean, covariance, names)

Sometimes the mean and covariance are given directly, e.g. in a publication, and you would define them in Python code as shown here, or read them from a file.

However, often you obtain these values as the result of a fit of a parametrised model to data, or estimate them in some other way.

Further examples to create MultiNorm objects are here: Create

Read only

MultiNorm objects should be used read-only objects!

If you need to change something (mean, covariance, names), make a new object!

TODO: make read-only as much as possible, the document remaining caveats!


Once you have a MultiNorm object representing a multivariate normal distribution, you can access the following properties and methods to analyse it.

The object repr only shows the number of dimensions (number of parameters) n of the distribution:

>>> mn

To see the contents, print the object:

>>> print(mn)
names: ['a', 'b', 'c']
mean: [10. 20. 30.]
error: [1. 2. 3.]
[[1. 0. 0.]
 [0. 4. 0.]
 [0. 0. 9.]]

You can access the attributes like this:

>>> mn.n
>>> mn.mean
array([10., 20., 30.])
>>> mn.cov
array([[1., 0., 0.],
       [0., 4., 0.],
       [0., 0., 9.]])
>>> mn.names
['a', 'b', 'c']

The mean and covar are numpy.ndarray objects. To be as accurate as possible, we always cast to 64-bit float on MultiNorm initialisation and do all computations with 64-bit floating point precision, even if 32-bit float or integer numbers are passed in.

>>> type(mn.mean)
>>> mn.mean.dtype

The mean is a 1-dimensional array, and cov is a 2-dimensional array:

>>> mn.mean.shape
>>> mn.cov.shape
(3, 3)

Parameter error vector MultiNorm.error:

>>> mn.error
array([1., 2., 3.])

Precision matrix (the inverse covariance) MultiNorm.precision:

>>> mn.precision
array([[1.        , 0.        , 0.        ],
       [0.        , 0.25      , 0.        ],
       [0.        , 0.        , 0.11111111]])

Correlation matrix MultiNorm.correlation:

>>> mn.correlation
array([[1., 0., 0.],
       [0., 1., 0.],
       [0., 0., 1.]])

These are just the basic attributes and properties.

We continue with this example on the Analyse page and show how to really do some analysis with MultiNorm objects and methods.


Plot ellipse using MultiNorm.to_matplotlib_ellipse:

import matplotlib.pyplot as plt
mn2 = mn.marginal()

Further examples to plot MultiNorm objects are here: Plot

What next?

The Create, Analyse and Plot tutorial pages contain further examples. The Theory and References pages contain background information and definitions, as well as links to other documents and codes.

The full API documentation is here: MultiNorm. Note that you can click on “source” on the right for any method or property, and read the implementation to see what exactly it does. It’s usually a few lines of straightforward code using Python and Numpy, so reading the source is recommended.