Apologies in advance if this just annoys you, but the current presentation of the COVID situation (on mainstream media) is driving me mad. You probably don't believe me when I say I'd quite like to be proved wrong, but actually I don't like the feeling of being a stranger in a strange land, and quite like being a normal human being. So please prove me wrong by a counter argument not by just shouting at me.
The BBC News reported 1,340 new infections (for Nov 30). That is immediately misleading because the correct statement would have been 1,340 PCR tests have tested positive. The relation between Test Positive Results (TPRs) and infections can be called into question, validly, in my view as will become clear, I hope.
One fact conveniently omitted was the number of tests done which was 214,845 as per the Government Corona dashboard, so a TPR ratio of 0.624%. Now I can see eyes rolling when I mention False Positives, but if you are not prepared to take account of these, please don't claim to be seriously following the science.
The Government admits is has no reliable figure for this in reply to a specific question. Though not in itself an official statement a paper which is from a paper on a Gov web site states:
The UK operational false positive rate is unknown. There are no published studies on the operational false positive rate of any national COVID-19 testing programme.
An attempt has been made to estimate the likely false-positive rate of national COVID-19 testing programmes by examining data from published external quality assessments (EQAs) for RT-PCR assays for other RNA viruses carried out between 2004-2019 [7]. Results of 43 EQAs were examined, giving a median false positive rate of 2.3% (interquartile range 0.8-4.0%).
I have tried to get a latest update, but with no success. A Lancet paper of 20 Sept 2020 repeats the 0.8 - 4%. This is of great significance because once the False Positive Rate exceeds the TPR rate, results are widely held to be meaningless.
So even taking, not the median but the best case from existing studies (0.8% FPR) with such a huge test population you would get 1,728 positive results even if the actual number of infections was just 10. Nearly 100% False Positives. Even halve the FPR and you're small gang of 10 infected people will produce 860 false positives.
The only was to make these mass tests worthwhile is by using test which is virtually perfect as regards specificity (which is just 100%-FPR). So yes, if the test specificity is 99.94% then the results are correct. But a test with such accuracy is not available. You can choose to believe that PCR is that good, but you do not have a shred of evidence. It's a sort of act of faith.
And BTW that's before you get into the scandal of the outsourcing of PCR tests from certified NHS labs to private labs which have no registration and are staffed by under-qualified people. But that's a whole other story.
Is someone able to convince me that the BBC report was accurate? and more importantly, mass testing of those without symptoms is worthwhile?
ONLY READ FURTHER IF YOU WANT TO CHECK THE METHOD OF CALCULATION
Test population (P): 214,845
Actual prevalence: 10 actual infections out of 241,845 tested
Infection probability (I) simply the above as a probability 0.0000465
FNR (False Negative Rate) = 10% this has minimum effect on the results
FPR (False Positive Rate) = 0.8% this is the most optimistic based on research
There's a much simpler problem, which is that some people take multiple tests (for example, as the infection progresses). Thus total number of tests != total number of infections.
(There are already two QR codes needed when flying anywhere at the moment. This will just become a third. I find keeping them in Apple Wallet the easiest way. You don’t even need to open your phone).
Even without QR codes, people flying anywhere still need to carry their passport and boarding pass. Adding "proof of Covid vaccination" to the list of things to carry isn't so hard, and most people won't forget. (Like Alan, I always have everything on paper, in case I run down the battery on whatever device I have.)
Whereas people are a lot more likely to forget to carry proof of vaccination when they want to visit the shop or the cinema. In the US, it's normal to be expected to present ID for admission to a bar, but if your friend is buying you lunch / taking you to the cinema, there's no reason up to now for you to carry anything at all. Which means there's plenty of opportunity for "I've been vaccinated, but my proof is on my kitchen counter / in my other pants / in my car".
Businesses, not wishing to turn away paying customers in these precarious times, are going to want to believe the customer. Those same businesses are known to have a somewhat patchy record at preventing underage drinking - it's beyond belief that they'll be better at preventing Covid drinking.
On the other hand, they don't need to be perfect. An OK job is probably good enough.
This is of great significance because once the False Positive Rate exceeds the TPR rate, results are widely held to be meaningless.
Anteater, you appear to be equating 2 entirely different things. Is the false positive rate the number of TOTAL TESTS that are falsely positive, as you're suggesting? Or is the false positive rate the number of POSITIVE TESTS that are false?
Because the answer to that question alters your maths entirely.
If you're wrong about the definition of what a false positive actually is, as I strongly suspect you are, then the rest of your argument is completely wrong.
You're actually arguing that 99% percent of positive tests are false, not 4%. Because you're trying to argue that 4% of tests will inevitably be positive. I repeat, I simply don't think you're correctly defining what "false positive" means, because you're calculating on the basis of how many TESTS are false, not how many POSITIVE TESTS are false.
And part of the reason I think you're wrong is because when you define 'false positive' in the way that you do, 'false negative' doesn't make any sense either.
I think the 10% FNR is low, the figures I've seen suggest that it's closer to 30% for a single PCR test (if, on the other hand you take two tests a few days apart the false negative rate for the combination is much lower). So, ideally people should be taking two tests a few days apart to confirm they're not infected (at that moment in time). FNR for the rapid flow tests seems to be much higher.
So, if the 1340 positive tests are 1340 new infections (zero false positive for the moment) then the actual number of new infections in the tested population would be somewhere around 1700. That's a lot of people carrying the virus without knowing it, though mass screening does mean that some people who are infected but ignorant know they are and self-isolate who would otherwise be potentially infecting others.
Now, for the FPR. If it is 0.8% (or higher) I agree with your maths that you would get 1500+ false positives from a 200,000 random sampling of the population. Which, if correct means that none of the 1300 ish positives are likely to be genuine. The proof of this will be in a couple of weeks - if the 1340 new infections reported for yesterday are real infections then that will result in somewhere like 150 daily hospital admissions (assuming the 20,000 plus new cases a few weeks ago are genuine - the false positive rate at the top end of 3,000 ish become irrelevant - and lead to the approx. 1500 daily hospital admissions at present => about 10% of infections result in hospital admission), if they're false positives the hospital admission rate will fall to very low in a few weeks - I know a simplification because the time lag between infection and hospitalisation is variable.
The added complication is that the 200,000 ish tests are not random. Some are community mass testing, many are still targeted to people with symptoms. If someone has symptoms then the chances of a real positive are much higher, which supresses the FPR associated with a genuine random sampling (there's a bit of fancy stats that relates to the outcomes of tests if there's a bias towards testing people likely to be positive - someone with symptoms is far more likely to be infected than someone without and this biases things, though I admit my stats knowledge isn't sufficient to follow the argument). There have been small scale genuine random sampling (using the approaches of polling agencies to select people for testing), but most testing is non-random: people get tests because they have symptoms, work in environments where they might be more exposed, or suspect they might have been infected; people who are convinced they haven't been infected don't generally go and get themselves tested.
A key part of the reason I think your wrong is that you're trying to start with an "actual prevalence" and work from there: claiming that 10 actual sick people leads to 1728 false positives.
Well no, that's completely backwards thinking in that the whole damn point of calculations of false positives is that the first bit of information you have is the test results, not the actual prevalence. You don't work out the tests from the prevalence, you work out the prevalence from the tests.
If the false positive rate is 4%, to take a simple (and high number), what that means is that 1738 positive tests, around 69 people weren't actually sick. 1,669 of them really were sick.
But then of course you have to take into account the false negatives... if you postulate a 10% false negative rate then that doesn't mean 1 infection was missed. It means 20,000 were missed.
Rather turns your argument on its head. The clue is in claiming that a 10% false negative rate doesn't affect the results.
Apologies in advance if this just annoys you, but the current presentation of the COVID situation (on mainstream media) is driving me mad. You probably don't believe me when I say I'd quite like to be proved wrong, but actually I don't like the feeling of being a stranger in a strange land, and quite like being a normal human being. So please prove me wrong by a counter argument not by just shouting at me.
The BBC News reported 1,340 new infections (for Nov 30). That is immediately misleading because the correct statement would have been 1,340 PCR tests have tested positive. The relation between Test Positive Results (TPRs) and infections can be called into question, validly, in my view as will become clear, I hope.
Actually 13,430 people tested positive - that’s UK government language, for what it’s worth - which renders the rest of your arithmetic off target by a factor of 10, before we even start to work on the logic...
In addition to which, one could point out that proportional numbers heading into hospitals, ICUs and coffins have provided some hard evidence correlating with infection numbers.
To be fair @orfeo, reading @Anteater's link it does appear that the FPR is %age of all tests (they go on to say that on a day where the -ve rate is 98.6% the FPR logically cannot be higher than 1.4%).
However, the paper appears to be contrasting that theoretical upper bound with the 2.3% rate which is not the FPR of the Covid test, but of " examining data from published external quality assessments (EQAs) for RT-PCR
assays for other RNA viruses carried out between 2004-2019. Results of 43 EQAs were examined,
giving a median false positive rate of 2.3% (interquartile range 0.8-4.0%)."
As they go on to say, the FPR cannot be that high and is likely to be much lower. @Anteater's problem is not misunderstanding the FPR but ising a figure which is not the Covid FPR from a paper which specifically says the Covid FPR must be much lower than this.
We should also observe that if the FPR rate is high therefore most PR are FPRs, then we would not see the huge variations in %PR over time, nor would the hospitalisation and death rates follow the PR with, in the latter case, a two week lag. People don't generally die of being told they have Covid.
To be fair @orfeo, reading @Anteater's link it does appear that the FPR is %age of all tests (they go on to say that on a day where the -ve rate is 98.6% the FPR logically cannot be higher than 1.4%).
However, the paper appears to be contrasting that theoretical upper bound with the 2.3% rate which is not the FPR of the Covid test, but of " examining data from published external quality assessments (EQAs) for RT-PCR
assays for other RNA viruses carried out between 2004-2019. Results of 43 EQAs were examined,
giving a median false positive rate of 2.3% (interquartile range 0.8-4.0%)."
As they go on to say, the FPR cannot be that high and is likely to be much lower. @Anteater's problem is not misunderstanding the FPR but ising a figure which is not the Covid FPR from a paper which specifically says the Covid FPR must be much lower than this.
Either way, part of why I can tell the maths is nonsense is because I live in a country where many many thousands of people are being tested and yet the number of positive tests is around 10 a day.
Which according to @Anteater is logically impossible, unless somehow Australia has a completely different and far better testing regime than the UK. According to @Anteater then even a completely disease-free Australia ought to be getting positive results for hundreds if not thousands of people each day.
False positives undoubtedly exist. But they have to be rare, or else having single figure daily results in this country couldn't happen.
The argument that the United Kingdom results are low can only work if you explain why the figures in different countries are so vastly different. I feel quite comfortable in asserting that the reason that the figures in Australia are vastly different from the UK or the USA is not because we've secretly got some special testing regime that protects us from false positives, but because we've actually managed (partly through the sheer luck of physical isolation) to be protected from the virus.
And numbers didn't spike in Melbourne mid-year because Melbourne somehow ran out of special Australian testing kits and had to rely on crappy British ones. Numbers spiked in Melbourne because the virus was circulating in Melbourne.
I am trying very hard not to get to the point of yelling about this, but the logic behind the supposed maths is so incredibly bad it's not funny. And people have been throwing bad maths about this into social media since at least March.
Plus the notion that no serious journalist has raised questions about this, but the truth can be discovered by someone on Ship of Fools with a history of wanting to deny that the pandemic is a serious problem... okay I'm walking away from the keyboard now because it's genuinely making me angry the more I think about it.
I think the fact that there is a clear trend in positive test results is more instructive than trying to guess at the false positive rate. If the infection rate were low enough for the positive results to be primarily false positives we would simply see noise in the figures - day-to-day fluctuations with no clear trend. If you look at the graph you can see a very clear rise to a peak and the beginnings of a fall. That's simply not consistent with such low infection levels (neither, in fact, are the hospitalisation and death figures, but presumably in covid-"sceptic" world those are due to a bad 'flu that everyone missed because they were looking for covid).
The state of Victoria got 6,874 test results yesterday. Number of positives? Zero.
New South Wales had 6, 635 tests reported. No local cases. 5 cases in returning overseas travellers.
Victoria currently doesn't accept overseas flights.
Do you maths on that, @Anteater . Magic Australian coronavirus testing? Special kits we reserve for overseas travellers in quarantine hotels that we source from the UK instead of using our precious magic ones?
unless somehow Australia has a completely different and far better testing regime than the UK.
To be fair, that wouldn't be hard. The only reason not to call the UK testing regime a shambles is that medieval slaughterhouses would be insulted.
The unbelievable claim is that the UK government is biased towards being overcautious and overstating the scale of the problem. Absolutely everything points in the direction of the UK government being overoptimistic, of talking down the scale of the problem, and of taking too little action too late. As Arethosemyfeet suggests, the data has consistently followed the trends predicted by people who've said the government is taking too little action rather than the predictions of people who think the government is doing too much.
(Johnson indignantly maintains that appealing to predictions that have turned out to be correct is working from hindsight.)
unless somehow Australia has a completely different and far better testing regime than the UK.
To be fair, that wouldn't be hard. The only reason not to call the UK testing regime a shambles is that medieval slaughterhouses would be insulted.
The unbelievable claim is that the UK government is biased towards being overcautious and overstating the scale of the problem. Absolutely everything points in the direction of the UK government being overoptimistic, of talking down the scale of the problem, and of taking too little action too late. As Arethosemyfeet suggests, the data has consistently followed the trends predicted by people who've said the government is taking too little action rather than the predictions of people who think the government is doing too much.
(Johnson indignantly maintains that appealing to predictions that have turned out to be correct is working from hindsight.)
If people were not dying, suffering greatly and sustaining long-term damage, it would almost be funny.
unless somehow Australia has a completely different and far better testing regime than the UK.
To be fair, that wouldn't be hard. The only reason not to call the UK testing regime a shambles is that medieval slaughterhouses would be insulted.
I think that the lateral flow tests are being mostly reserved for healthcare and the university students "home for Christmas" plan, where testing is frequent enough it will help mitigate the FNR somewhat.
The local councils are hopping mad about being kept in Tier 3 even though the rates are dropping. Apparently it's because the government reckon not enough people are getting tested, and the positive rate for those who are is quite high. (It's certainly been doing the rounds of the local schools since half term.) I've been past our local test centre quite a few times recently, but only seen one person going in.
Apologies in advance if this just annoys you, but the current presentation of the COVID situation (on mainstream media) is driving me mad. You probably don't believe me when I say I'd quite like to be proved wrong, but actually I don't like the feeling of being a stranger in a strange land, and quite like being a normal human being. So please prove me wrong by a counter argument not by just shouting at me.
The BBC News reported 1,340 new infections (for Nov 30). That is immediately misleading because the correct statement would have been 1,340 PCR tests have tested positive. The relation between Test Positive Results (TPRs) and infections can be called into question, validly, in my view as will become clear, I hope.
Actually 13,430people tested positive - that’s UK government language, for what it’s worth - which renders the rest of your arithmetic off target by a factor of 10, before we even start to work on the logic...
All that math ... made worthless by a factor of 10 error? It's not like I've never slipped a decimal, so perhaps some charity is in order. On the other hand, I suspect this number is reported regularly (it certainly is in the US) so you'd think that someone paying attention would notice an order of magnitude error.
And if you're looking at the ratio of "people tested positive"/"virus tests conducted" (a bit dodgy, since some people are tested multiple times) and comparing it to the FPR, look back to data from the summer - August 1, for example:
Tests conducted: 146,011
People tested positive: 543
By Anteater's reckoning this establishes a FPR no higher than 543/146,011=0.372%. Applying this rate to Tuesday's test total of 214,835 (not 214,845) we would get an upper bound of some 799 false positives, or about 5.95% of the 13,430 positive cases.
I think the fact that there is a clear trend in positive test results is more instructive than trying to guess at the false positive rate. If the infection rate were low enough for the positive results to be primarily false positives we would simply see noise in the figures - day-to-day fluctuations with no clear trend.
This passed without comment but it's a key point. Even the 'noise' in the data we have isn't noisy - the hair on the UK curves has shape due to weekly-cyclical patterns in testing and reporting. The noise floor is very low - in August we could have had a go at measuring it by testing people more than once and doing some averaging, but I guess we had better things to do with those tests.
Is the false positive rate the number of TOTAL TESTS that are falsely positive, as you're suggesting? Or is the false positive rate the number of POSITIVE TESTS that are false?
Four result parameters are defined:
True Positive: Condition exists and test is positive.
True Negative: Condition does not exist and test is negative.
False Positive: Condition does not exist but test is positive
False Negative: Condition exists but test is positive.
Two test quality parameters are defined, confusingly in two ways:
Sensitivity is the probability that the condition will be detected when it exists, and FNR = 1-sensitiviy
Specificity is the probability that the condition is absent and the test will be negative, and FPR = 1-specificity.
So as to your question, the FPR is neither of the numbers, it is a probability not a number. When I stated that in the example, nearly 100% were False Positives that is not the same as claiming a FPR for the test of nearly 100%.
I do indeed argue that 99+% of the positives are false and I claim that this is shown by the maths assuming a FPR = 0.8%, the most optimistic value based on research I have found. Similarly I argue that to get to the result of 1,340 actual infections you would need an FPR of 0.06%, which I believe to be way beyond the current test capability.
I agree that the terminology gets very confusing. Saying that 99% of the positives were false is very close to saying that the test's false positive rate is 99%. So clarity is probably helped by having a moratorium on the term false positive rate.
The BBC News reported 1,340 new infections (for Nov 30). That is immediately misleading because the correct statement would have been 1,340 PCR tests have tested positive. The relation between Test Positive Results (TPRs) and infections can be called into question, validly, in my view as will become clear, I hope.
Actually 13,430people tested positive - that’s UK government language, for what it’s worth - which renders the rest of your arithmetic off target by a factor of 10, before we even start to work on the logic...
All that math ... made worthless by a factor of 10 error? It's not like I've never slipped a decimal, so perhaps some charity is in order. On the other hand, I suspect this number is reported regularly (it certainly is in the US) so you'd think that someone paying attention would notice an order of magnitude error.
If one starts a maths-based argument by misquoting a number it is not going to end well. Especially if the intent is to downplay the scale of infections painfully evident to the rest of us from correlated hospital admissions and deaths. On the same day, 1369 hospital admissions and 603 deaths were reported.
As for the rest of your remarks, I refer to the excellent points already made by @orfeo -
The state of Victoria got 6,874 test results yesterday. Number of positives? Zero.
New South Wales had 6, 635 tests reported. No local cases. 5 cases in returning overseas travellers.
Victoria currently doesn't accept overseas flights.
Do your maths on that, @Anteater . Magic Australian coronavirus testing? Special kits we reserve for overseas travellers in quarantine hotels that we source from the UK instead of using our precious magic ones?
The Pfizer vaccine has been approved by the U.K. regulator, roll out begins next week.
I'm curious as to how various skeptics will react, and how many will veer into anti-vax territoty. Twitter already awash with anti-vax sentiment.
They will react by getting a fever, and/or a change in their sense of taste or smell, and/or a new continuous cough. After that they will either: recover; progress to serious illness, hospitalization and in some cases die; or acquire a chronic condition that may not resolve for months.
So clarity is probably helped by having a moratorium on the term false positive rate.
Why? It's clearly defined term (you linked to a Wiki page defining it), though there are questions about what value to assign to it in any particular case. We can make educated estimates of limits to that value. As @orfeo reported recent tests in two jurisdications in Australia were over 6000 tests in each, with zero cases. We can realistically therefore say that in 10,000 tests we can expect no more than 1 false positive - that gives an upper limit to the FPR of 1/10,000 = 0.01%. At that limit you'd be seeing 20 (maximum) false positives in the 200,000 plus tests, and 14,000 plus true positives, we're talking about. Which we'd probably all like to be lower, but it's not a great problem - those are 20 people who are self isolating for a couple of weeks who probably didn't need to, but there's no harm in that ... there's a good argument that we should all be reducing our social contact anyway (those 20 people have a couple of weeks when they're less likely to contract coronavirus from someone else).
Of much greater concern is the known fact that the false negative rate is quite high, which means that there are people who have been tested in our communities with coronavirus who think they're OK because the test was negative. You're example was FNR=10% (and, I've never seen any report of a FNR that low, 30% seems more commonly reported), and if we take that then if x is the number of false negatives (people with the virus and a negative test) and we have 14,000 true positives as per the 30th Nov figures we get:
0.1 = x/(x+14000) (the number of condition positives is the number for got a +ve result + the unknown people who are infected but got a -ve result)
0.1x +1400 = x
0.9x = 1400
x = 1555
So, even with a low FNR that's over 1500 people walking around, potentially infecting others, who we don't know from the cohort of those who were tested, a small fraction of the population (but probably not representative, so I won't scale that by the total population to estimate the number of unknown carriers in the country).
The Pfizer vaccine has been approved by the U.K. regulator, roll out begins next week.
I'm curious as to how various skeptics will react, and how many will veer into anti-vax territoty. Twitter already awash with anti-vax sentiment.
They will react by getting a fever, and/or a change in their sense of taste or smell, and/or a new continuous cough. After that they will either: recover; progress to serious illness, hospitalization and in some cases die; or acquire a chronic condition that may not resolve for months.
Well, no, if lots of people have the vaccine, those who don't will be protected. Then the skeptics will say, see, I told you I don't need a vaccine.
I think the 10% FNR is low, the figures I've seen suggest that it's closer to 30% for a single PCR test . . . FNR for the rapid flow tests seems to be much higher.
I suspect your right about FNR, but the calculations I used are not at all sensitive to it, so I didn't bother being accurate here. The figures quote on PHE for the LFT are 76.8% and a higher specificity of 99.68%. If the tests had used LFT and if the specificity is really 99.68% you'd still get 687 FPS for 7 TPs which shows how hard it is to accurately test huge populations.
So, if the 1340 positive tests are 1340 new infections (zero false positive for the moment) then the actual number of new infections in the tested population would be somewhere around 1700.
True but as we say "If pigs could fly, pork would go up". I really wish we had perfect tests but that's just not the case.
Now, for the FPR. If it is 0.8% (or higher) I agree with your maths . . . The proof of this will be in a couple of weeks - if the 1340 new infections reported for yesterday are real infections then that will result in somewhere like 150 daily hospital admissions . . .
I agree that the data on admissions is key. And if the data on COVID admissions referred to people being admitted because they really were ill with COVID then I would agree. But that's not how the statistics are compiled. The explanatory document accompanying the admission statistic for England states:
For all relevant data items: a confirmed COVID-19 patient is any patient admitted to the trust who has recently (ie in the last 14 days) tested positive for COVID-19 following a polymerase chain reaction (PCR) test.
This is in line with the WHO definition of a confirmed case, which requires no symptoms at all but only a PCR positive test.
So if any of the false positive people in the mass test, breaks a leg and is admitted, they are admitted as a COVID case, and count towards the total COVID admissions, despite the true probability of their being infected being vanishingly small.
Also, I have been informed (from a skeptic source - so biassed, but if my view totally believable) that people on admission are routinely tested for COVID. There is a lot of evidence for this on NHS sites, e.g. from UC Hospital London:
COVID-19 - patient testing
If you are admitted to our hospital, either overnight, or for day surgery, you will be tested to see if you have COVID-19. This includes patients who may not have any symptoms of COVID-19.
So again, if I break my leg, I go to hospital and get a PCR test despite having no symptoms, and if positive I am now a COVID patient.
I do not believe that the NHS records statistics specifically for those who have been admitted because they are showing symptoms of COVID, so I am of the opinion that mass testing with huge numbers of FPs are leading to exaggerated numbers of COVID admissions. The admissions are real, it's the connection to COVID that is questionable.
The added complication is that the 200,000 ish tests are not random.
True, but I do not see this as relevant. Nobody I know of, who could broadly described as skeptical of the current UK measures, dismisses the testing done in the Spring, and indeed they support the use of PCR as a confirmation accompanying a medical diagnosis based on symptoms. These can be very much more accurate because of the obvious higher prevalence in the population of people with a fever, cough etc. The objection is to using a test with a significant number of false results as a diagnosis in the absence of any symptoms.
The Pfizer vaccine has been approved by the U.K. regulator, roll out begins next week.
I'm curious as to how various skeptics will react, and how many will veer into anti-vax territoty. Twitter already awash with anti-vax sentiment.
They will react by getting a fever, and/or a change in their sense of taste or smell, and/or a new continuous cough. After that they will either: recover; progress to serious illness, hospitalization and in some cases die; or acquire a chronic condition that may not resolve for months.
Well, no, if lots of people have the vaccine, those who don't will be protected. Then the skeptics will say, see, I told you I don't need a vaccine.
I don't really care how many people don't get it, or their reasons for making that decision. Just as long as enough people do get it for this bloody lockdown to finally end.
Alternatively, if the government goes with a "you're free to do what you want just as long as you've got Proof Of Vaccination in your wallet" policy then I really don't care how many refuse it. Let the freaks, idiots and paranoiacs keep themselves locked away rather than accept a vaccination. No skin off my nose. I'll be out there having fun and going back to living my life again while they carry on staring at the same four walls they've been staring at since March.
The BBC News reported 1,340 new infections (for Nov 30). That is immediately misleading because the correct statement would have been 1,340 PCR tests have tested positive. The relation between Test Positive Results (TPRs) and infections can be called into question, validly, in my view as will become clear, I hope.
Actually 13,430people tested positive - that’s UK government language, for what it’s worth - which renders the rest of your arithmetic off target by a factor of 10, before we even start to work on the logic...
All that math ... made worthless by a factor of 10 error? It's not like I've never slipped a decimal, so perhaps some charity is in order. On the other hand, I suspect this number is reported regularly (it certainly is in the US) so you'd think that someone paying attention would notice an order of magnitude error.
If one starts a maths-based argument by misquoting a number it is not going to end well. Especially if the intent is to downplay the scale of infections painfully evident to the rest of us from correlated hospital admissions and deaths. On the same day, 1369 hospital admissions and 603 deaths were reported.
I'm entirely in agreement with you, Cameron - the "someone paying attention" I was referring to was @Anteater, who apparently still has not noticed the glaring factor of 10 error.
The objection is to using a test with a significant number of false results as a diagnosis in the absence of any symptoms.
@orfeo has addressed the flaws in that argument with Australian data. Your calculations palpably fail to connect with hard data. In addition, as @Dave W points out, you still don't seem to have noticed the factor of 10 error in your calculations which started you off on this line of argument.
...huge numbers of FPs are leading to exaggerated numbers of COVID admissions.
No, that is obviously wrong. People aren't admitted to hospital in the UK based on a COVID test. They are admitted when the disease progresses to the point that their condition can't be managed at home. At that point they not only have a positive test, but a lot of other clinical indicators to confirm the nature of the disease.
When a positive test for duck genes comes with quacking, feathers and a fondness for pond life then any chance of spurious debate about it being a duck is over.
The description in the UK statistics is the number of "Covid19 patients". The obvious interpretation of that is people who are being treated for Covid19. It would seem bizarre to refer to someone being treated for cancer or a broken leg as a Covid19 patient just because they also have a positive test result for coronavirus (though, if they also develop covid-19 symptoms they could be classed as both a cancer patient and a covid19 patient). You need symptoms to to be counted as having covid-19 ... merely being infected by the virus isn't sufficient.
Well, obviously you need to test everyone admitted to hospital (and staff, visitors and others) because you need to do your best to prevent coronavirus spread among patients - who, almost by definition are among the most vulnerable. And, a hospital will need information on which patients are infected to manage resources such as deployment of PPE. But, it still looks odd to classify someone who is infected with the virus but not showing symptoms of the disease, or even someone only showing mild symptoms, as a covid-patient. It's not the most odd thing ever though.
I think functionally, it probably not possible to process the data at that level of detail fast enough to do the separation, report and then nationally aggregate.
Some amusing tweets appearing, e.g., "I will accept a vaccine if you can prove that US voting machines were not hacked by the Clintons, and 5G cell towers are dismantled, and Chinese labs take back the virus from the paedophile pizza ring, then I'll be satisfied, and only then".
So again, if I break my leg, I go to hospital and get a PCR test despite having no symptoms, and if positive I am now a COVID patient.
I do not believe that the NHS records statistics specifically for those who have been admitted because they are showing symptoms of COVID, so I am of the opinion that mass testing with huge numbers of FPs are leading to exaggerated numbers of COVID admissions.
Count the dead people. Deaths track hospital admissions, with the expected time delay. And hospital admissions track test positivity rate. With time delays, obviously.
Without doing anything, this tells you that the rate of hospital admissions and the test results are giving you roughly the right picture. Sure - we can do some math and quantify that, but just by looking, you see that the story is roughly consistent.
Slightly late to reply, as washing egg of the face takes some time. Captain Mainwaring would definitely say: "I'm pleased to see you spotted my deliberate mistake". Still it helped me, and may have given a bit of mood-raising schadenfreude, so drinks all round. At least someone said
It's not like I've never slipped a decimal, so perhaps some charity is in order.
In fact there is some mileage in this, and I still stick to a few points:
1. Don't report positive results as infections
2. I still think I can demonstrate the PCR, or even LFT testing on random populations is worthless.
3. I think it is a pity that the facts are hard to find.
I have a breakdown of the tests in the various tiers, and rejigged the model to take account of this. Tier 1 is defined as
“Virus testing in Public Health England (PHE) labs and NHS hospitals for those with a clinical need, and health and care workers
So I tried to see how the figures worked using a Specificity of 99.2% as before and guessing the prevalence in Tier 1.
And the data all match based on a guess prevalence of 25% in the Tier 1 population. You can still play around with lower figures for the Tier 1 prevalence and lower figures for Specificity. But at the upper bound error bound (.977) it works with 14% Tier 1 prevalence, although here, about a third of the positives are false.
I've no idea of what the real figure is but it is clear that it is a dominating factor. If only the report giving the number of tests in each tiers also gave the number of positives for each tier. But no such luck.
So I'll retire from this discussion with as much dignity as I can and am glad to have got a better picture.
In fact there is some mileage in this, and I still stick to a few points:
1. Don't report positive results as infections
2. I still think I can demonstrate the PCR, or even LFT testing on random populations is worthless.
3. I think it is a pity that the facts are hard to find.
Not that hard to find the number was 13,430 rather than 1,343.
I've no idea of what the real figure is but it is clear that it is a dominating factor.
Bullshit. Using your own method, we can see that even if every positive during the summer was a false positive, only about 6% of today’s positives would be false.
You were off by a factor of 10, still have “no idea of what the real figure is”, but yet somehow you’re still convinced of the correctness of your original position. Not very intellectually dignified.
2. I still think I can demonstrate the PCR, or even LFT testing on random populations is worthless.
The evidence of the trial in Liverpool would contradict your assertion. Mass testing identifies people who are infected who would otherwise be unknown (or, not known until symptoms develop a few days later). Providing those infected people isolate, then every infected person who is no longer infecting others is a bonus. Of course, we'd all like the tests to be better so that less people who are infected walk out thinking they're OK. But, we'll take what we can get.
The only potential counter argument is if people who test negative (some of whom will be infected) significantly change behaviour and stop social distancing because they think they're OK. I've not seen any evidence of that happening on a large scale, but happy to be corrected if that has happened.
The evidence of the trial in Liverpool would contradict your assertion.
I agree and I withdraw it.
The only true statement is that the any test can produce unreliable results as the presumed prevalence approaches the false positive rate, which I think is widely accepted.
My estimate is that a prevalence of 3.6% in Liverpool (3 time the UK average which sounds reasonable) leaves you with 10% FPs, which goes up to 50% for a prevalence of 0.4%.
So I was proved wrong. Well I did apologise in advance in case I wasted your time, but thanks for your contributions.
Next time I try this, I'll make sure it's April 1st, so I'll have an excuse.
Next time I try this, I'll make sure it's April 1st, so I'll have an excuse.
April Fool's Day is a valid reason for spreading dangerous disinformation during a lethal pandemic? What do you do for a encore, shout 'Fire!' in a crowded theatre?
I commend the Brits for approving the Pfizer vaccine. I imagine the Moderna Vaccine will be approved shortly as well. Poor Donald. He wanted to get it approved here before anyone else.
I actually think the Brits will help us understand the rollout and will give us an idea of how the vaccine will be accepted, and eventually what it will take to flatten and reduce the curve.
If the Rfactor gets down to near 0 by August I propose we all celebrate at a London pub. Or at least a pub in our own regions.
I believe that's true of the AstraZeneca, Pfizer, and Moderna vaccines.
The Pfizer and Moderna vaccines require two shots a couple of weeks apart, and need sophisticated cold stores (the Moderna one is not as reliant on dry ice temperature storage), which does create logistically problems on top of those of most other vaccines - so it's not just the same as the 'flu jab needing to reach many more people.
The AstraZeneca vaccine works with a single shot at about 60-70% effectiveness, the second shot is needed to boost that to the 80%+ range, and can be stored and distributed in normal fridge temperatures. That makes the logistics of a vaccination programme to give decent levels of immunity much simpler - but, especially in circumstances where a significant proportion of the population refuses to get the jab, the second shot would be needed to rapidly suppress community transmission and allow lifting of social distancing restrictions.
There will be people out there working on the logistics and which vaccines and approaches would be most suitable to particular situations. As a moderately informed non-expert if you're in an area with low incidence of coronavirus then a mass vaccination programme with the single-shot AstraZeneca would be enough to allow re-opening, with a second shot or using one of the other vaccines for those individuals who are particularly vulnerable or exposed. In areas of higher incidence we're probably going to see a greater use of double shot AstraZeneca with a later re-opening (both because it'll take longer to get people double dosed, plus you'll need to get a larger proportion of the population vaccinated). In richer nations, the Pfizer and Moderna vaccines will be the preference for high risk individuals because of the higher effectiveness, but the cheaper to produce and distribute AstraZeneca (or other vaccines which will be coming along) will be the most commonly used. The more expensive and difficult to handle vaccines may be almost exclusively used in richer nations.
Comments
The BBC News reported 1,340 new infections (for Nov 30). That is immediately misleading because the correct statement would have been 1,340 PCR tests have tested positive. The relation between Test Positive Results (TPRs) and infections can be called into question, validly, in my view as will become clear, I hope.
One fact conveniently omitted was the number of tests done which was 214,845 as per the Government Corona dashboard, so a TPR ratio of 0.624%. Now I can see eyes rolling when I mention False Positives, but if you are not prepared to take account of these, please don't claim to be seriously following the science.
The Government admits is has no reliable figure for this in reply to a specific question. Though not in itself an official statement a paper which is from a paper on a Gov web site states:
I have tried to get a latest update, but with no success. A Lancet paper of 20 Sept 2020 repeats the 0.8 - 4%. This is of great significance because once the False Positive Rate exceeds the TPR rate, results are widely held to be meaningless.
So even taking, not the median but the best case from existing studies (0.8% FPR) with such a huge test population you would get 1,728 positive results even if the actual number of infections was just 10. Nearly 100% False Positives. Even halve the FPR and you're small gang of 10 infected people will produce 860 false positives.
The only was to make these mass tests worthwhile is by using test which is virtually perfect as regards specificity (which is just 100%-FPR). So yes, if the test specificity is 99.94% then the results are correct. But a test with such accuracy is not available. You can choose to believe that PCR is that good, but you do not have a shred of evidence. It's a sort of act of faith.
And BTW that's before you get into the scandal of the outsourcing of PCR tests from certified NHS labs to private labs which have no registration and are staffed by under-qualified people. But that's a whole other story.
Is someone able to convince me that the BBC report was accurate? and more importantly, mass testing of those without symptoms is worthwhile?
ONLY READ FURTHER IF YOU WANT TO CHECK THE METHOD OF CALCULATION
Test population (P): 214,845
Actual prevalence: 10 actual infections out of 241,845 tested
Infection probability (I) simply the above as a probability 0.0000465
FNR (False Negative Rate) = 10% this has minimum effect on the results
FPR (False Positive Rate) = 0.8% this is the most optimistic based on research
True Positives: P*I*(1-FNR)=214845*0.0000465*.9 = 9
True negatives P*(1-I)*(1-FPR) =214845* 0.9999535*(1-0.008)= 213,106
False Negatives P*I*FNR=214845*0.0000465*.1=1
False Positives P*(1-I)*FPR =214845* 0.9999535*0.008= 1,728
Even without QR codes, people flying anywhere still need to carry their passport and boarding pass. Adding "proof of Covid vaccination" to the list of things to carry isn't so hard, and most people won't forget. (Like Alan, I always have everything on paper, in case I run down the battery on whatever device I have.)
Whereas people are a lot more likely to forget to carry proof of vaccination when they want to visit the shop or the cinema. In the US, it's normal to be expected to present ID for admission to a bar, but if your friend is buying you lunch / taking you to the cinema, there's no reason up to now for you to carry anything at all. Which means there's plenty of opportunity for "I've been vaccinated, but my proof is on my kitchen counter / in my other pants / in my car".
Businesses, not wishing to turn away paying customers in these precarious times, are going to want to believe the customer. Those same businesses are known to have a somewhat patchy record at preventing underage drinking - it's beyond belief that they'll be better at preventing Covid drinking.
On the other hand, they don't need to be perfect. An OK job is probably good enough.
Anteater, you appear to be equating 2 entirely different things. Is the false positive rate the number of TOTAL TESTS that are falsely positive, as you're suggesting? Or is the false positive rate the number of POSITIVE TESTS that are false?
Because the answer to that question alters your maths entirely.
If you're wrong about the definition of what a false positive actually is, as I strongly suspect you are, then the rest of your argument is completely wrong.
You're actually arguing that 99% percent of positive tests are false, not 4%. Because you're trying to argue that 4% of tests will inevitably be positive. I repeat, I simply don't think you're correctly defining what "false positive" means, because you're calculating on the basis of how many TESTS are false, not how many POSITIVE TESTS are false.
And part of the reason I think you're wrong is because when you define 'false positive' in the way that you do, 'false negative' doesn't make any sense either.
So, if the 1340 positive tests are 1340 new infections (zero false positive for the moment) then the actual number of new infections in the tested population would be somewhere around 1700. That's a lot of people carrying the virus without knowing it, though mass screening does mean that some people who are infected but ignorant know they are and self-isolate who would otherwise be potentially infecting others.
Now, for the FPR. If it is 0.8% (or higher) I agree with your maths that you would get 1500+ false positives from a 200,000 random sampling of the population. Which, if correct means that none of the 1300 ish positives are likely to be genuine. The proof of this will be in a couple of weeks - if the 1340 new infections reported for yesterday are real infections then that will result in somewhere like 150 daily hospital admissions (assuming the 20,000 plus new cases a few weeks ago are genuine - the false positive rate at the top end of 3,000 ish become irrelevant - and lead to the approx. 1500 daily hospital admissions at present => about 10% of infections result in hospital admission), if they're false positives the hospital admission rate will fall to very low in a few weeks - I know a simplification because the time lag between infection and hospitalisation is variable.
The added complication is that the 200,000 ish tests are not random. Some are community mass testing, many are still targeted to people with symptoms. If someone has symptoms then the chances of a real positive are much higher, which supresses the FPR associated with a genuine random sampling (there's a bit of fancy stats that relates to the outcomes of tests if there's a bias towards testing people likely to be positive - someone with symptoms is far more likely to be infected than someone without and this biases things, though I admit my stats knowledge isn't sufficient to follow the argument). There have been small scale genuine random sampling (using the approaches of polling agencies to select people for testing), but most testing is non-random: people get tests because they have symptoms, work in environments where they might be more exposed, or suspect they might have been infected; people who are convinced they haven't been infected don't generally go and get themselves tested.
Well no, that's completely backwards thinking in that the whole damn point of calculations of false positives is that the first bit of information you have is the test results, not the actual prevalence. You don't work out the tests from the prevalence, you work out the prevalence from the tests.
If the false positive rate is 4%, to take a simple (and high number), what that means is that 1738 positive tests, around 69 people weren't actually sick. 1,669 of them really were sick.
But then of course you have to take into account the false negatives... if you postulate a 10% false negative rate then that doesn't mean 1 infection was missed. It means 20,000 were missed.
Rather turns your argument on its head. The clue is in claiming that a 10% false negative rate doesn't affect the results.
Actually 13,430 people tested positive - that’s UK government language, for what it’s worth - which renders the rest of your arithmetic off target by a factor of 10, before we even start to work on the logic...
In addition to which, one could point out that proportional numbers heading into hospitals, ICUs and coffins have provided some hard evidence correlating with infection numbers.
However, the paper appears to be contrasting that theoretical upper bound with the 2.3% rate which is not the FPR of the Covid test, but of " examining data from published external quality assessments (EQAs) for RT-PCR
assays for other RNA viruses carried out between 2004-2019. Results of 43 EQAs were examined,
giving a median false positive rate of 2.3% (interquartile range 0.8-4.0%)."
As they go on to say, the FPR cannot be that high and is likely to be much lower. @Anteater's problem is not misunderstanding the FPR but ising a figure which is not the Covid FPR from a paper which specifically says the Covid FPR must be much lower than this.
We should also observe that if the FPR rate is high therefore most PR are FPRs, then we would not see the huge variations in %PR over time, nor would the hospitalisation and death rates follow the PR with, in the latter case, a two week lag. People don't generally die of being told they have Covid.
Either way, part of why I can tell the maths is nonsense is because I live in a country where many many thousands of people are being tested and yet the number of positive tests is around 10 a day.
Which according to @Anteater is logically impossible, unless somehow Australia has a completely different and far better testing regime than the UK. According to @Anteater then even a completely disease-free Australia ought to be getting positive results for hundreds if not thousands of people each day.
False positives undoubtedly exist. But they have to be rare, or else having single figure daily results in this country couldn't happen.
The argument that the United Kingdom results are low can only work if you explain why the figures in different countries are so vastly different. I feel quite comfortable in asserting that the reason that the figures in Australia are vastly different from the UK or the USA is not because we've secretly got some special testing regime that protects us from false positives, but because we've actually managed (partly through the sheer luck of physical isolation) to be protected from the virus.
And numbers didn't spike in Melbourne mid-year because Melbourne somehow ran out of special Australian testing kits and had to rely on crappy British ones. Numbers spiked in Melbourne because the virus was circulating in Melbourne.
I am trying very hard not to get to the point of yelling about this, but the logic behind the supposed maths is so incredibly bad it's not funny. And people have been throwing bad maths about this into social media since at least March.
Plus the notion that no serious journalist has raised questions about this, but the truth can be discovered by someone on Ship of Fools with a history of wanting to deny that the pandemic is a serious problem... okay I'm walking away from the keyboard now because it's genuinely making me angry the more I think about it.
The state of Victoria got 6,874 test results yesterday. Number of positives? Zero.
New South Wales had 6, 635 tests reported. No local cases. 5 cases in returning overseas travellers.
Victoria currently doesn't accept overseas flights.
Do you maths on that, @Anteater . Magic Australian coronavirus testing? Special kits we reserve for overseas travellers in quarantine hotels that we source from the UK instead of using our precious magic ones?
The unbelievable claim is that the UK government is biased towards being overcautious and overstating the scale of the problem. Absolutely everything points in the direction of the UK government being overoptimistic, of talking down the scale of the problem, and of taking too little action too late. As Arethosemyfeet suggests, the data has consistently followed the trends predicted by people who've said the government is taking too little action rather than the predictions of people who think the government is doing too much.
(Johnson indignantly maintains that appealing to predictions that have turned out to be correct is working from hindsight.)
I think that the lateral flow tests are being mostly reserved for healthcare and the university students "home for Christmas" plan, where testing is frequent enough it will help mitigate the FNR somewhat.
The local councils are hopping mad about being kept in Tier 3 even though the rates are dropping. Apparently it's because the government reckon not enough people are getting tested, and the positive rate for those who are is quite high. (It's certainly been doing the rounds of the local schools since half term.) I've been past our local test centre quite a few times recently, but only seen one person going in.
And if you're looking at the ratio of "people tested positive"/"virus tests conducted" (a bit dodgy, since some people are tested multiple times) and comparing it to the FPR, look back to data from the summer - August 1, for example:
Tests conducted: 146,011
People tested positive: 543
By Anteater's reckoning this establishes a FPR no higher than 543/146,011=0.372%. Applying this rate to Tuesday's test total of 214,835 (not 214,845) we would get an upper bound of some 799 false positives, or about 5.95% of the 13,430 positive cases.
This passed without comment but it's a key point. Even the 'noise' in the data we have isn't noisy - the hair on the UK curves has shape due to weekly-cyclical patterns in testing and reporting. The noise floor is very low - in August we could have had a go at measuring it by testing people more than once and doing some averaging, but I guess we had better things to do with those tests.
True Positive: Condition exists and test is positive.
True Negative: Condition does not exist and test is negative.
False Positive: Condition does not exist but test is positive
False Negative: Condition exists but test is positive.
Two test quality parameters are defined, confusingly in two ways:
Sensitivity is the probability that the condition will be detected when it exists, and FNR = 1-sensitiviy
Specificity is the probability that the condition is absent and the test will be negative, and FPR = 1-specificity.
So as to your question, the FPR is neither of the numbers, it is a probability not a number. When I stated that in the example, nearly 100% were False Positives that is not the same as claiming a FPR for the test of nearly 100%.
I do indeed argue that 99+% of the positives are false and I claim that this is shown by the maths assuming a FPR = 0.8%, the most optimistic value based on research I have found. Similarly I argue that to get to the result of 1,340 actual infections you would need an FPR of 0.06%, which I believe to be way beyond the current test capability.
It's well explained here on Wikipedia.
I agree that the terminology gets very confusing. Saying that 99% of the positives were false is very close to saying that the test's false positive rate is 99%. So clarity is probably helped by having a moratorium on the term false positive rate.
Anteater made the error, so I don’t know what you mean by “you think that someone paying attention would notice.” The BBC did not report the lower figure Anteater mentioned at all.
If one starts a maths-based argument by misquoting a number it is not going to end well. Especially if the intent is to downplay the scale of infections painfully evident to the rest of us from correlated hospital admissions and deaths. On the same day, 1369 hospital admissions and 603 deaths were reported.
As for the rest of your remarks, I refer to the excellent points already made by @orfeo -
I'm curious as to how various skeptics will react, and how many will veer into anti-vax territoty. Twitter already awash with anti-vax sentiment.
They will react by getting a fever, and/or a change in their sense of taste or smell, and/or a new continuous cough. After that they will either: recover; progress to serious illness, hospitalization and in some cases die; or acquire a chronic condition that may not resolve for months.
Of much greater concern is the known fact that the false negative rate is quite high, which means that there are people who have been tested in our communities with coronavirus who think they're OK because the test was negative. You're example was FNR=10% (and, I've never seen any report of a FNR that low, 30% seems more commonly reported), and if we take that then if x is the number of false negatives (people with the virus and a negative test) and we have 14,000 true positives as per the 30th Nov figures we get:
0.1 = x/(x+14000) (the number of condition positives is the number for got a +ve result + the unknown people who are infected but got a -ve result)
0.1x +1400 = x
0.9x = 1400
x = 1555
So, even with a low FNR that's over 1500 people walking around, potentially infecting others, who we don't know from the cohort of those who were tested, a small fraction of the population (but probably not representative, so I won't scale that by the total population to estimate the number of unknown carriers in the country).
Well, no, if lots of people have the vaccine, those who don't will be protected. Then the skeptics will say, see, I told you I don't need a vaccine.
So if any of the false positive people in the mass test, breaks a leg and is admitted, they are admitted as a COVID case, and count towards the total COVID admissions, despite the true probability of their being infected being vanishingly small.
Also, I have been informed (from a skeptic source - so biassed, but if my view totally believable) that people on admission are routinely tested for COVID. There is a lot of evidence for this on NHS sites, e.g. from UC Hospital London: So again, if I break my leg, I go to hospital and get a PCR test despite having no symptoms, and if positive I am now a COVID patient.
I do not believe that the NHS records statistics specifically for those who have been admitted because they are showing symptoms of COVID, so I am of the opinion that mass testing with huge numbers of FPs are leading to exaggerated numbers of COVID admissions. The admissions are real, it's the connection to COVID that is questionable. True, but I do not see this as relevant. Nobody I know of, who could broadly described as skeptical of the current UK measures, dismisses the testing done in the Spring, and indeed they support the use of PCR as a confirmation accompanying a medical diagnosis based on symptoms. These can be very much more accurate because of the obvious higher prevalence in the population of people with a fever, cough etc. The objection is to using a test with a significant number of false results as a diagnosis in the absence of any symptoms.
I don't really care how many people don't get it, or their reasons for making that decision. Just as long as enough people do get it for this bloody lockdown to finally end.
Alternatively, if the government goes with a "you're free to do what you want just as long as you've got Proof Of Vaccination in your wallet" policy then I really don't care how many refuse it. Let the freaks, idiots and paranoiacs keep themselves locked away rather than accept a vaccination. No skin off my nose. I'll be out there having fun and going back to living my life again while they carry on staring at the same four walls they've been staring at since March.
@orfeo has addressed the flaws in that argument with Australian data. Your calculations palpably fail to connect with hard data. In addition, as @Dave W points out, you still don't seem to have noticed the factor of 10 error in your calculations which started you off on this line of argument.
No, that is obviously wrong. People aren't admitted to hospital in the UK based on a COVID test. They are admitted when the disease progresses to the point that their condition can't be managed at home. At that point they not only have a positive test, but a lot of other clinical indicators to confirm the nature of the disease.
When a positive test for duck genes comes with quacking, feathers and a fondness for pond life then any chance of spurious debate about it being a duck is over.
From a hospital perspective it matters because they will still need to barrier nurse you.
Count the dead people. Deaths track hospital admissions, with the expected time delay. And hospital admissions track test positivity rate. With time delays, obviously.
Without doing anything, this tells you that the rate of hospital admissions and the test results are giving you roughly the right picture. Sure - we can do some math and quantify that, but just by looking, you see that the story is roughly consistent.
In fact there is some mileage in this, and I still stick to a few points:
1. Don't report positive results as infections
2. I still think I can demonstrate the PCR, or even LFT testing on random populations is worthless.
3. I think it is a pity that the facts are hard to find.
I have a breakdown of the tests in the various tiers, and rejigged the model to take account of this. Tier 1 is defined as So I tried to see how the figures worked using a Specificity of 99.2% as before and guessing the prevalence in Tier 1.
And the data all match based on a guess prevalence of 25% in the Tier 1 population. You can still play around with lower figures for the Tier 1 prevalence and lower figures for Specificity. But at the upper bound error bound (.977) it works with 14% Tier 1 prevalence, although here, about a third of the positives are false.
I've no idea of what the real figure is but it is clear that it is a dominating factor. If only the report giving the number of tests in each tiers also gave the number of positives for each tier. But no such luck.
So I'll retire from this discussion with as much dignity as I can and am glad to have got a better picture.
You were off by a factor of 10, still have “no idea of what the real figure is”, but yet somehow you’re still convinced of the correctness of your original position. Not very intellectually dignified.
The only potential counter argument is if people who test negative (some of whom will be infected) significantly change behaviour and stop social distancing because they think they're OK. I've not seen any evidence of that happening on a large scale, but happy to be corrected if that has happened.
I agree and I withdraw it.
The only true statement is that the any test can produce unreliable results as the presumed prevalence approaches the false positive rate, which I think is widely accepted.
My estimate is that a prevalence of 3.6% in Liverpool (3 time the UK average which sounds reasonable) leaves you with 10% FPs, which goes up to 50% for a prevalence of 0.4%.
So I was proved wrong. Well I did apologise in advance in case I wasted your time, but thanks for your contributions.
Next time I try this, I'll make sure it's April 1st, so I'll have an excuse.
April Fool's Day is a valid reason for spreading dangerous disinformation during a lethal pandemic? What do you do for a encore, shout 'Fire!' in a crowded theatre?
I actually think the Brits will help us understand the rollout and will give us an idea of how the vaccine will be accepted, and eventually what it will take to flatten and reduce the curve.
If the Rfactor gets down to near 0 by August I propose we all celebrate at a London pub. Or at least a pub in our own regions.
I wish the UK and our UK Shipmates all the best with the vaccine.
That may well be. I haven't dug into the details. I figure that may take a while to get sorted out. I just happened to hear about the UK version.
~~
In case anyone's interested, here's some of what's going on in California and San Francisco. And--oh!--there are some stats to play with!
"[California Gov.] Newsom says stay-at-home order likely if COVID-19 surge continues" (SF Gate).
"San Francisco sees 265% increase in COVID test positivity: What you need to know" (SF Gate, a branch of the SF Chronicle).
Here in SF, we just went back to a curfew (10 pm-8 am, IIRC).
The AstraZeneca vaccine works with a single shot at about 60-70% effectiveness, the second shot is needed to boost that to the 80%+ range, and can be stored and distributed in normal fridge temperatures. That makes the logistics of a vaccination programme to give decent levels of immunity much simpler - but, especially in circumstances where a significant proportion of the population refuses to get the jab, the second shot would be needed to rapidly suppress community transmission and allow lifting of social distancing restrictions.
There will be people out there working on the logistics and which vaccines and approaches would be most suitable to particular situations. As a moderately informed non-expert if you're in an area with low incidence of coronavirus then a mass vaccination programme with the single-shot AstraZeneca would be enough to allow re-opening, with a second shot or using one of the other vaccines for those individuals who are particularly vulnerable or exposed. In areas of higher incidence we're probably going to see a greater use of double shot AstraZeneca with a later re-opening (both because it'll take longer to get people double dosed, plus you'll need to get a larger proportion of the population vaccinated). In richer nations, the Pfizer and Moderna vaccines will be the preference for high risk individuals because of the higher effectiveness, but the cheaper to produce and distribute AstraZeneca (or other vaccines which will be coming along) will be the most commonly used. The more expensive and difficult to handle vaccines may be almost exclusively used in richer nations.
The one that has been approved in the U.K. is two vaccinations 21 days apart, then 10 days after that you will be protected.
I really don’t understand the thinking behind curfews. The virus is just as virulent during the day as after 10pm. 🤫
Absolutely - and also then at the bewitching hour, there is a wealth of folk returning home filling up public transport and the streets!