Reporting delays

[ywteh]

Use simple idea in https://epiforecasts.io/covid/posts/national/united-kingdom/

Can we deal with weekend/holiday effects?

[mhutchin-isolabs]

The current delay profile probably is not as we would like it. Currently, we use a Gamma(5, 1) distribution. Unfortunately, this gives a mean just under the mean serial interval, and a median delay of about 3 days and does not account for random test delay.

The suggestion is to switch to something with a tighter variance and account for individual test delay as well as the serial interval. Suggestions include:

• Tightening the variance on the Gamma distribution for the serial interval.
• Switching the Gamma for lognormal for the serial interval.
• Adding 1 + Poisson(1 or 2) as a random test delay after noticing symptoms.