ArviZ supportΒΆ

As of version 2.2, corner supports plotting ArviZ samples, providing a consistent interface to many MCMC samplers.

Here is a simple example of this:

import corner
import numpy as np
import arviz as az

np.random.seed(11234)

x = np.random.randn(2, 2000)
data = az.from_dict(
    posterior={
        "x": x,
        "y": np.random.randn(2, 2000, 2)
    },
    sample_stats={
        "diverging": x < -1.2
    }
)

figure = corner.corner(data, divergences=True)
../_images/arviz_1_0.png

When using this interface, it is probably best to provide the truths parameter using a dictionary keyed on the variable names:

figure = corner.corner(data, truths={"x": 1.5, "y": [-0.3, 0.1]})
../_images/arviz_3_0.png