Inference with bayesian network in PyMC3

Questions version agnostic
2

Hi @SCon.
So I think that the thread you link to (and Pomegranate) are a distraction. I believe everything you want to do here can be done in PyMC.

It really sounds like you are describing Signal Detection Theory, so you are basically doing psychophysics here :slight_smile:

Take a look at Implementing normal CDF for signal detection theory model (includes full sample code).

Note that in signal detection theory they use terms like

  • “hit rate” which is the same as a “correct detection” (1-alpha) and
  • “miss rate” which is the same as a Type II error / false negative (beta)

See Type I and type II errors - Wikipedia for more on this if it’s not clear.

Feel very free to shout if this doesn’t solve your problem.

1 Like