Table 2.

Parameters and prior distributions for the initial window used for Bayesian inference. These prior distributions are only used at the start and are not used in the rest of the sequential inference, where in each window, the prior is an over dispersed version of the posterior in the previous window (see Sub-Section 3.5). The prior for $\beta$ is a diffuse long-tail log normal centered at $\beta =1$. The prior for $g$ and $\omega$ are nearly uniform, also non-informative, but avoiding the unexpected a prior values of zero and one, Capistran et al. (2021).

Parameter	Prior distribution
Contact rate $\beta$	$log\left(\beta \right)\sim N(0,1)$
Fraction of infected dying ($g$)	$Beta(1+1/6,1+1/3)$
Proportion of the effective population ($\omega$)	$Beta(1+1/6,1+1/3)$