calthaus / ncov-cfr

R code and documentation to estimate case fatality ratio of COVID-19 from cases outside China
GNU General Public License v3.0
90 stars 17 forks source link

Distribution of onset-to-death time in the Linton et al database #6

Open tamas-ferenci opened 4 years ago

tamas-ferenci commented 4 years ago

You seem to fit a gamma distribution to the raw data pulled from the Linton et al study, but my understanding is that we should rather use estimates that are adjusted for right truncation (i.e., Table 2). It means a substantial difference, first, because the best fitting curve is log-normal and not gamma (actually it is true even for the raw numbers of onset-to-death in Table 1, at least based on AIC), and perhaps even more importantly, there is a huge difference in the dispersion: your estimates for the gamma distribution imply an SD of 6.7, while Table 2 gives 11.6 (or 12.7 if admission-to-death is considered).