Modelling is dodgy, some say. But if we fit the real published data to it, using data to adjust model parameters, we increase confidence in the model, perhaps? THREAD full plot in thread 1/n
#COVIDー19 #COVID19

Executive summary (details further down the thread). Key values are:
R0 = 2.72 used to give match
% of population already infected: 41-50% (23-28 million, England only data used) 2/n

Plot shows
- Modelled infection numbers (red)
- Modelled infections (cumulative), including recovered (black)
- Published data for cumulative infection numbers, scaled (blue)
- Number of active infections (orange)

3/n

Model is pretty straightforward (see link at end of thread), based on homogenous growth based on R0 and population size. Social distancing/lockdown neasures modelled as a weighting for R at specific dates. Cumulative number of infections used is the raw published data 4/n

Current infections is obtained from the sum of 7 consecutive days of new daily cases,
assuming average number of days for recovery is seven.
Variables are day 0, R0, weightings used for R on 16 Mar and 23 Mar.
5/n

Plot also shows scenario in which restrictions released on 1 May but this is after reference date so has no effect on fitting, of course. Also, published data scaled to match model since model shows all infections not tested only 6/n

Using this model fitted to the real data, on reference date (20 April) we have between 23-28 million already infected (herd immunity at 33-36 million). R0 used for this fit is 2.72 7/n

Read more at (start from the bottom!) https://paper.dropbox.com/doc/Personal-brain-dump-of-thoughts-on-Covid-19-pandemic--AycNdxkF1UdcyLJ~S25M1WJxAQ-2uwTsFM5EJbdnZwvYHd1R
Above document updated from time to time /END OF THREAD/