Add .dist().plot() method to pymc distributions

[ezDataCode]

It would be nice to quickly visualize a pymc distribution. This video and this notebook describes how it can be done. Since this is a common situation with a known solution, I think this feature should be added as a built-in .dist().plot() method of all pymc distributions. Any reason not to add this as a standard feature?

[ricardoV94]

The API since PyMC3 (which that example uses). Something like pm.plot_dist(rv) would make more sense. Replace self.draw (which no longer exists) in that example by pm.draw(rv, draws=1000)

[aloctavodia]

No strong opinion about this just wanted to mention two points.

This idea was already proposed and rejected, mainly argument was something along the lines of not (re)adding plotting code to pymc. Of course this can be revisited.

In the meantime, or as alternative you can use preliz to visualize distributions. It has methods for pdf, cdf and ppf, and also an interactive method to change parameters with sliders. https://preliz.readthedocs.io/en/latest/examples/param_space_1d_examples.html

[ricardoV94]

Does preliz give a mapping between its parametrizations and the ones PyMC uses?

[aloctavodia]

PreliZ uses the same parametrization used by PyMC, including alternative parametrization.