Problem creating new random variable class

Questions
4

  1. I only had a quick glance of the Miller distribution, so I am not completely sure what is this slots you are referring to. If you meant that P, Q, R, S are themselves distributions, you can pass a distribution to it, eg: pm.Normal.dist(mu=0., sd=1.,). Similar usage could be found in LKJCholeskyCov: http://docs.pymc.io/api/distributions/multivariate.html#pymc3.distributions.multivariate.LKJCholeskyCov

  2. If you implemented a plot function, probably the best way to access it within a model context is var.distribution.plot().

  3. Yes. And if you are outside of a model context you can do MillerDistribution.dist(...).logp(). eg: pm.Normal.dist(0., 1.).logp(0.)

It would be great to extend that page, contribution welcome! :wink:

1 Like