Read on Twitter
The MRC mid-point forecast (made on 12th Oct) was for 465 deaths / day by 26 Oct & a range of 240-690.
PHE data are now fairly complete for 26 Oct & actual total is 201. A few more will be reported later but final total seems likely to be well below even the lower bound forecast https://twitter.com/cricketwyvern/status/1319971794716102659
PHE data are now fairly complete for 26 Oct & actual total is 201. A few more will be reported later but final total seems likely to be well below even the lower bound forecast https://twitter.com/cricketwyvern/status/1319971794716102659
Forecasting is really difficult & I don’t want to criticise individuals. But it is important that lessons are learnt.
This forecast was made for a date only 2 weeks ahead, i.e. it was based on infections already in the system (given the lag between infections & deaths).
This forecast was made for a date only 2 weeks ahead, i.e. it was based on infections already in the system (given the lag between infections & deaths).
Also, the forecast range was very wide (240 to 690) so it is very striking that data has come in below even the lower bound.
No wonder then that people are sceptical about the modelling the Government presented on Saturday suggesting up to 4,000 deaths per day.
No wonder then that people are sceptical about the modelling the Government presented on Saturday suggesting up to 4,000 deaths per day.
Throughout the crisis, there seems to be a systematic tendency for forecasts & modelling to predict very scary numbers which rarely come to pass.
The Government slides supporting the latest lockdown presented a range of scenarios but every single one has daily deaths significantly above the peak in April and reaching that point in less than 2 months time.
These predictions are made at a time when infections have stabilised (from new positives/Zoe App estimates).
Who knows if it will continue & nobody can be certain about future deaths. But is it reasonable not to present a single scenario in which deaths peak at less than April?
Who knows if it will continue & nobody can be certain about future deaths. But is it reasonable not to present a single scenario in which deaths peak at less than April?
The worry many people have is that modelling/forecasts are being used in a political way to promote lockdown decisions rather than to present the public with a reasonable range of plausible outcomes.
