Computing condition probability in a Bayesian Network with data

I have fitted parameters of a Bayesian Network from data. The results of the marginal probability are similar to when I use another package that uses a maximum likelihood approach.

However, when I try computing prob (veh_type|injury = 1) or prob (night|injury = 1) by assigning 1 value at observed for injury line in a code, (similar to other code I have seen on different discussions) the results are not correct.

Need help on how to compute condition probabilities. I am interested to compute prob (injury|night = 1) or prob (injury|veh_type = 1) and diagnostic inference, prob (veh_type|injury = 1). I appreciate your help.


Hi. I think this reply and generally the whole thread should be very interesting to you: Bayes Nets, Belief Networks, and PyMC

1 Like