Starting over here with NCP and rescaled to relative shedding variability * mean genomes shed (where mean genomes shed is estimated in log scale).
Additionally, I took Adrian's suggestion and am now using modified versions of his lognormal3 functions which parameterize the lognormal with the log of the mean (not the $mu$!) and the scale parameter in unit scale, so that $MC(t)$ is the mean not the median and sigma is on the unit scale.
Able to get only a few divergences with a relative diffuse prior on the cv in individual shedding, so this seems like progress! Going to test out running this on all locations for our comparison forecast date to see how it performs...
Attempt to address #51
Starting over here with NCP and rescaled to relative shedding variability * mean genomes shed (where mean genomes shed is estimated in log scale).
Additionally, I took Adrian's suggestion and am now using modified versions of his
lognormal3functions which parameterize the lognormal with the log of the mean (not the $mu$!) and the scale parameter in unit scale, so that $MC(t)$ is the mean not the median andsigmais on the unit scale.Able to get only a few divergences with a relative diffuse prior on the
cvin individual shedding, so this seems like progress! Going to test out running this on all locations for our comparison forecast date to see how it performs...