Thread by @GrahamNeary

Some observations on the faulty infection fatality rate assumptions which drove the push towards lockdown.

Ferguson's Imperial College paper, advocating lockdown, assumed an infection fatality rate in the UK of 0.9%.

Its source for this assumption was another paper by Ferguson.

This other paper admitted that there wasn't very good data available to estimate the threat posed by Covid-19.

To estimate IFR, you need to estimate how many infections there have been.

The best way to do that is with a randomised antibody test. Or as the paper said:

Since they didn't have an antibody test available, they used something else to estimate total infections.

They used the test for active infection (not the test for antibodies, or prior infection) that was used on people being repatriated from Wuhan. This is the PCR test.

That sounds like it might be useful, right?

It turns out that the test involved 689 individuals on six flights.

Six people tested positive:

These six people are the basis of estimating the number of infections in Wuhan.

And therefore the basis of estimating Wuhan's infection fatality rate.

Which was used to model the UK's IFR.

Which resulted in the estimate of 500k deaths in the UK.

Predicting 500,000 UK deaths (or 2.2 million US deaths) using 6 positive cases as your input data might seem odd.

But there's more.

The test gives an unknown number of false negatives. It's useless when there is no active infection.

TLDR: Ferguson used the output data from an unreliable test on a tiny sample size, which he admitted was not the right sort of test to begin with, to come up with the scenarios in which 500k British people and 2.2 million Americans died.