How to keep dimensions in a posterior predictive sample?

Questions v5
4

Someone more clever than I can chime in with simpler solutions, but I would do one of these;

  • use the values of mu in your posterior to parameterize 2 Poisson distributions and use those to generate credible observations associated with each event type
  • group your posterior predictions by event type and average all the posterior predictive samples in each group

But like I said, there may a quicker, more concise solution that I can’t think of at the moment.