For the miscellaneous Rt estimation notebook, the choice of tau (sliding window width) = 6 seems to affect the way the poisson and negative binomial models differ, by causing them to experience different posterior mean Rt at some regions of time. A shorter sliding window seems to lead to the models having more similar posterior means, so that the difference between the models is primarily differing posterior uncertainty. Maybe the differing posterior means is interesting for some reason and should be investigated? Otherwise, a shorter sliding window width for this problem could be preferable.
For the miscellaneous Rt estimation notebook, the choice of tau (sliding window width) = 6 seems to affect the way the poisson and negative binomial models differ, by causing them to experience different posterior mean Rt at some regions of time. A shorter sliding window seems to lead to the models having more similar posterior means, so that the difference between the models is primarily differing posterior uncertainty. Maybe the differing posterior means is interesting for some reason and should be investigated? Otherwise, a shorter sliding window width for this problem could be preferable.