Obviously, the reason I ask this question is that a theme is going around some of the Usual Suspects that since there is a FPR of 1%, the rise in cases seen since August is meaningless, we're being panicked into losing our liberties ...1/ coronavirus.data.gov.uk/?_ga=2.1128524…
https://twitter.com/Gilesyb/status/1307937303801802757
But even if the people sampled are entirely randomly chosen, the implications of the rise of cases from <1000 to >4000 (from 150k tests) is surely significant - and they are not randomly chosen (unless you think the actual prevalence was practically zero at the end of August) 2/
(in fact, if there have been a steady 1500 people reported as +ve who are -ve who should be cut from the figures, the rise from practically nothing to 2500 in three weeks sounds even more worrying) 3/
But presumably the people sampled are not random, and have a much higher likelihood of being truly positive, making that 1% FPR less relevant to the analysis, or in need of modification?
After all, 4000 from ~150,000 people tested is a 2.67% prevalence or about 1 in 40 ... 4/
After all, 4000 from ~150,000 people tested is a 2.67% prevalence or about 1 in 40 ... 4/
whereas what I read is that the best estimate right now is 1 in 900 bbc.co.uk/news/uk-517682…
Which is pretty high! 5/
Which is pretty high! 5/
Bottom line:
Trust the epidemiologists
They have heard of "False Positive" and know how to adjust
The trend in the numbers is what counts.
6/6
(thanks for your patience as I think aloud...)
Trust the epidemiologists
They have heard of "False Positive" and know how to adjust
The trend in the numbers is what counts.
6/6
(thanks for your patience as I think aloud...)
• • •
Missing some Tweet in this thread? You can try to
force a refresh
