Read on Twitter
*peeps head around Twitter-door*
Oh, the notifications have gone quiet after the swarm of Peter Hitchens fanboys moved on. Excellent. Wonder what statistical abuses Twitter will have moved on to now...
Journalists: COVID TESTS HAVE A 90% FALSE POSITIVE RATE! AHHH!!!!!!
Me:
Oh, the notifications have gone quiet after the swarm of Peter Hitchens fanboys moved on. Excellent. Wonder what statistical abuses Twitter will have moved on to now...
Journalists: COVID TESTS HAVE A 90% FALSE POSITIVE RATE! AHHH!!!!!!
Me:
First, "false positive rate" and "false discovery rate" aren't the same.
The "false positive rate" is the number of false positives out of the total number of negatives.
The "false discovery rate" is the number of false positives out of the total number of positives
The "false positive rate" is the number of false positives out of the total number of negatives.
The "false discovery rate" is the number of false positives out of the total number of positives
So unless you know exactly which one you're talking about - and I mean *know*, not just seen someone on twitter talking about it - you can't even begin to try to draw conclusions.
Second, if it is the "false positive rate" you're talking about, it's still NOT 90% ffs
Second, if it is the "false positive rate" you're talking about, it's still NOT 90% ffs
The idea is that if the "false positive rate" is, say 1%, and you do 100,000 tests, then you will find 1000 'false positives'.
And so if you only found like 1,111 cases in total, that would mean 90% of the cases you found could have been false positives
And so if you only found like 1,111 cases in total, that would mean 90% of the cases you found could have been false positives
THAT IS NOT THE SAME AS A TEST HAVING A 90% FALSE POSITIVE RATE
And it's not even clear if the 1% figure is the 'false positive rate' or the 'false discovery rate' as most journalists use this term interchangeably despite them being totally different things
And it's not even clear if the 1% figure is the 'false positive rate' or the 'false discovery rate' as most journalists use this term interchangeably despite them being totally different things
If it's 1% 'false discovery rate' then it means that if you find 100 positives then 1 is likely to be false. Big deal.
Even if it was the "90%" example above, that's not much consolation to the people in Madrid who have just had their scheduled operations cancelled again because they've had to convert rooms into emergency ICUs and set up tents outside as their hospitals have reached saturation
*bursts in to new, make-shift emergency ICU*
"Don't worry, chaps! When incidence was really low, the proportion of false positives was pretty high!"
Yeah thanks for that mate, really helpful.
Has no bearing on actual number of people going in to hospital needing treatment
"Don't worry, chaps! When incidence was really low, the proportion of false positives was pretty high!"
Yeah thanks for that mate, really helpful.
Has no bearing on actual number of people going in to hospital needing treatment
And, AND! The "false" number would be based on testing a random bunch of people. Not a self selecting sample of people with symptoms
Think of the self selecting twitter polls you see showing Jeremy Corbyn was going to get 80% of the vote.
That's how much sampling changes stuff
Think of the self selecting twitter polls you see showing Jeremy Corbyn was going to get 80% of the vote.
That's how much sampling changes stuff
Really struggling to cope with the statistical mutilations going on during this crisis, not gunna lie
