A corner plotting routine for MCMC output in python. Requires numpy and matplotlib.
If you already have numpy and matplotlib, then install the package using pip by
running
pip install https://github.com/anguswilliams91/CornerPlot/archive/master.zip
The plotting function is also called corner_plot and has an informative docstring.
Below is an example output, where samples from a unit, isotropic Gaussian are plotted.
import corner_plot as cp
import numpy as np
#draw samples from a unit, isotropic Gaussian
samples = np.random.multivariate_normal([10.,5.,100.],[[1.,0.,0.],[0.,1.,0.],[0.,0.,1.]],size=1000000)
#plot
axis_labels=["$x$","$y$","$z$"]
cp.corner_plot(samples,axis_labels=axis_labels)
cp.corner_plot(samples,gradient=True,linewidth=0.,nbins=100,axis_labels=axis_labels)
cp.corner_plot(samples[::1000],scatter=True,filled=False,scatter_size=2,axis_labels=axis_labels,nbins=10)
If you have constrained the same model with different data sets, you can now plot the inference from each
data set on the same corner plot to compare the results. This is implemented in a function multi_corner_plot.
Example:
import corner_plot as cp
import numpy as np
samples = np.random.multivariate_normal([10.,21.,101.],[[1.,0.,0.],[0.,1.,0.],[0.,0.,1.]],size=1000000)
samples1 = np.random.multivariate_normal([12.,20.,100.],[[1.,0.,0.],[0.,1.,0.],[0.,0.,1.]],size=1000000)
samples2 = np.random.multivariate_normal([9.,22.,98.],[[1.,0.,0.],[0.,1.,0.],[0.,0.,1.]],size=1000000)
chains = (samples,samples1,samples2)
cp.multi_corner_plot(chains,axis_labels=["$x$","$y$","$z$"],linewidth=2.,\
chain_labels=["data 1","data 2","data 3"])
