[ ] Make a forwards model with negative binomial distribution (use sensible parametrization and document it carefully)
[ ] Make a notebook where you generate data from this, and try to learn Rt using the existing poisson based posterior. Show that as overdsipersion increases, it gets worse