Note

Click here to download the full example code

Multiple Truth Values

By default truth values are plotted with horizontal and vertical lines. However, if you have multiple truth values, this causes intersections which have no meaning.

This can be alleviated visually by ensuring your different truth values are in different colours or line styles, but if you end up having numerous truth values, the grid of intersecting lines can still become an eyesore.

See the below for an example.

from chainconsumer import Chain, ChainConsumer, Truth

cols = ["A", "B"]
chain1 = Chain.from_covariance([0.0, 0.0], [[1.0, -1], [-1, 2]], columns=cols, name="Contour 1")
chain2 = Chain.from_covariance([3.0, 3.0], [[1.0, 1], [1, 2]], columns=cols, name="Contour 2")
c = ChainConsumer()
c.add_chain(chain1)
c.add_chain(chain2)
c.add_truth(Truth(location={"A": 0, "B": 0}, color="#fb7185"))
c.add_truth(Truth(location={"A": 3, "B": 3}, color="#1f2937"))
fig = c.plotter.plot()

To fix this, you have two options:

1. Add markers instead of truth values.
2. Tell the truth values to plot using a marker.

To do option 2, simple pass in marker to the Truth object. This will ensure that your marginalised plots still have vertical lines at the truth values, while your contour plots will have markers.

You also can pass in marker_edge_width and marker_size to control the appearance of the markers. Note though that matplotlib doesn't let you have marker edges for all markers, but if you try to increase the line width on a marker which doesn't support it, ChainConsumer will raise an error so you're not scratching your head over why the plots aren't changing!

c = ChainConsumer()
c.add_chain(chain1)
c.add_chain(chain2)
c.add_truth(Truth(location={"A": 0, "B": 0}, color="#fb7185", marker="o"))
c.add_truth(Truth(location={"A": 3, "B": 3}, color="#1f2937", marker="X"))
fig = c.plotter.plot()

The above example will give you points on the subplots showing 2D slices of the marginalised posterior, and vertical lines in the 1D marginalised histograms. If this makes the histograms too noisy and you only want to see the points in the contour plots, consider swapping over to simply putting in markers.

I also apologise for the inconsistent naming here - marker becomes marker_style. Sorry.

c = ChainConsumer()
c.add_chain(chain1)
c.add_marker(
    location={"A": 0, "B": 0},
    name="Truth1",
    color="#fb7185",
    marker_style="o",
)
c.add_marker(
    location={"A": 3, "B": 3},
    name="Truth2",
    color="#1f2937",
    marker_style="X",
)
fig = c.plotter.plot()

Total running time of the script: ( 0 minutes 2.607 seconds)

Download Python source code: plot_5_multiple_truth_values.py

Download Jupyter notebook: plot_5_multiple_truth_values.ipynb

Gallery generated by mkdocs-gallery