Figure out why sampling for Z uses Q and Pi

evidation-health / ContinuousTimeMarkovModel

Continuous-time Markov model with discrete observations

MIT License

23 stars 23 forks source link

Figure out why sampling for Z uses Q and Pi #7

Closed pourzanj closed 9 years ago

pourzanj commented 9 years ago

When Z is sampled the log-likelihood for Q and Pi is invoked, when these have no direct relation with Z. This hints at a bug in the comorbidity likelihood.

pourzanj commented 9 years ago

It doesn't actually "use" these value per se.This was just an artifact of how theano prints the logpt of the model