Generation time uncertainty

[seabbs]

I see use of a fixed generation time which would lead to spuriously precise results. I'm not aware of any rationale to support this assumption as it seems unlikely to be exact across space and time (even if estimated with some degree of certainty). It would be great to clarify this as it can be a major contributor to uncertainty (to the extent we have started using multiple generation time estimates as sensitivity analyses).

LocalCovidTracker/prep/daily_tracker_scraper.R

Line 93 in f029651

utla.R.backcalculated <- estimate_R(utla.incidence,

See here for a walkthrough of some of the uncertain options and here for an implementation of the non-parametric approach.