normal_corner

This code produces a corner plot for analytical multi-dimensional Gaussian distribution, using covariance matrix and mean matrix. It also allows us to plot another distribution, with reduced dimensionality, on top. I.e. a distribution for a case where we fixed one variable.

Examples

It's as simple as figure_1 = normal_corner.normal_corner(covariance_matrix,mean_vector,variable_labels)!

See demo.py for detailed usage examples.

Installation

Option 1:

pip install normal_corner

Option 2:

python setup.py install in cloned directory

Documentation

The main component is a normal_corner function inside a normal_corner package.

def normal_corner(covm,mean,varlabels,fixedvarindex=None,fixedvarvalue=None,
           covm2=None,mean2=None,scale_factor=3,diagnostic=False,
           color='red',color2='blue', fig=None, **fig_kw):

Below is a description of inputs and outputs.

Output:

Matplotlib figure object with a corner plot

Main input:

covm : covariance matrix, numpy array, NxN.
mean : mean matrix, numpy array, 1xN.
varlabels : labels for plotting, 1xN, list of str in LaTex format, between ($$).

Input for a second distribution on top:

fixedvarindex : index of variable that we do not use (fix), int, starting from 0. If not None, covm2 and mean2 must not be None.
fixedvarvalue : value of fixed variable, float. Leave None not to plot fixed value.
newcov and newmean : new covariance and mean matrices, same format, as above.

Optional input:

scale_factor : scale factor for plotting area, float, in sigma.
diagnostic : an option to print out some diagnostic messages, bool.
color : color for a main Normal distribution, str.
color2 : color for a secondary Normal distribution, with reduced dimensionality, str.
fig : Matplotlib figure to plot to, possibly output of corner.corner to plot on top of MCMC corner plot
additional keywords are passed to figure().

bvgoncharov/normal_corner