Pymc3 can output some pretty spiffy-looking DAG diagrams using the function pm.model_to_graphviz:
(this is a recreation of the errors-in-variables bit from ch14 of “Statistical Rethinking”
And not only do they look nice, they’re also super useful for people like me who still have a bit of trouble understanding the model specification syntax (esp with multiple likelihoods). The diagrams help me be sure Pymc3 is understanding what I want!
Anyway it’s a shame this feature is only mentioned like once in the tutorial examples. I think more people should know about it, and one logical place to put it would be in “How to debug a model”. Inspecting the DAG visualization to see if it looks like the model you have in mind should be one of the first things you try if things aren’t working as they should.
