Good point. 97% accuracy isn't great.

My province is declaring a pre-emptive state of emergency tomorrow. We have about 21 cases in the province, and the local government wants some legal standing to issue a curfew...

God I hope this cluster-fuck ends soon..

Coming soon to a sane government near you. — Baden

That order of "You are not allowed to go out for a walk for fresh air" is simply stupid. (Especially when your annoying neighbor with an ugly dog can go.)

The only logical reason would be that Spaniards don't give a damn about regulations and hence it's easier for the police to break up people socializing in the park or whatever. Yet making the rules more draconian that "as people disobey them partly" would be enough is simply doesn't work.

Wouldn't it be nice if we were vastly over-estimating the occurrence? If the tests give a false positive 3% of the time, and we rapidly increase the number of people we're testing, it would look like a rapid increase in infections during that time..

If only...

That order of "You are not allowed to go out for a walk for fresh air" is simply stupid. (Especially when your annoying neighbor with an ugly dog can go.) — ssu

HEY! My little Quasimodo is NOT ugly!

That's the only one I really don't like as I exercise a lot. Mostly running. But... anyone can just say "Hey, I was just going for a walk", so if you allow that it has the potential to more or less makes a nonsense of the lockdown.

This really goes down to things like the common sense and how people take authorities and orders from the government.

For example, cafes and restaurants haven't been closed here, but they have seen a loss of 90% of their customers. And people do keep the distance when walking. The simple fact is that when you have only 1 death in the country it's hard to issue very draconian regulations. And if the vast majority will obey, is that enough to curb the peak and have the health sector functioning? After all, the reason for the lockdown is to have it effect the people in a longer time period. Time will tell.

Using your numbers (and assuming they apply to England as a whole).

164/20883 = .00785.

A test that's 97% accurate will be wrong 3% of the time. False positives, then, 1.5%.

.015 x 20883 = 313

That means if you test positive, you're chances are about 164/313 of being positive = about 50-50.

But that applies to your mum only if her self-diagnosis is 97% accurate. Very likely she's got her symptoms right, but since less than 1% of the population is positive (evidently), then why suppose that her diagnostic skills are even remotely near her ability to list symptoms? The P of her being positive based on her self-diagnostician's skills falls off very rapidly as her accuracy decreases.

people will abuse the loophole it creates as the Italians did and why they're in a stricter lock down.

Thanks to Trump, the USA seems to be winning the corona Olympics. Now in third place.

No harm no foul. He was just being optimistic. :shade:

No harm no foul for Senator Burr either. Just being optimistic with the public while privately telling his friends there was a disaster coming and selling hundreds of thousands of dollars in hotel and other stocks. Politicians like Trump and Burr just didn't want the good people of America to worry their silly little heads about things while they looked after their own interests. Nothing to see here.

I just read an article on The Atlantic about, correct me if wrong, a stimulus package for Danish to continue work in the private sector for 75 percent in original salary pay for 13 weeks, which would cost quite a lot.

It's a shoddy article, but inspiration is due for such a plan.

https://www.google.com/amp/s/amp.theatlantic.com/amp/article/608533/

Thanks to Trump, the USA seems to be winning the corona Olympics. Now in third place. — Benkei

Actually the US is way down the list in infections per capita. So let's do this, since it's all a game of sorts to you, should the final tally show the US with a lower per capita infection and death rate than The Netherlands, would you be willing to admit to Trump's superiority in handling this crisis to The Netherlands? Or, is it like Trump's age old malaria drug hunch, where we don't need additional proof of what the evidence will be, we already just know?

Or, is it like Trump's age old malaria drug hunch, where we don't need additional proof of what the evidence will be, we already just know? — Hanover

You must be high. The age old malaria drug, which is chloroquine or Hydroxychloroquine are effective, per reports of South Korea and China...

The French have come to the same conclusion, and if I'm not mistaken the Canadians are going to test it on 1,500 subjects (not humans)...

If you think I think it's a game, you haven't read my posts.

My friend tested me on that a couple of weeks ago:

Out of 20,338 people tested in Britain for covid-19 164 people have the disease. The test itself is 97% accurate. You take the test and it comes back positive. What's the chance you actually have it based on this single test?

Apparently the answer is something like 21%? — Michael

Assuming you got the prevalence of the disease right, which is the difficult part of the calculation.

If we assume the base chance to have the virus is 0.8 percent, as per your numbers, then out of a thousand people 8 will have the virus. These 8 will test positive (rounded up). But from the remaining 992 another 30 will test positive (3%, rounded up). So if you test positive, your chance of having the virus are 8/38, which is 21%.

Conclusion: for rare diseases, you need very accurate tests.

Teams of experts could easily be assembled to tell the headless chickens what to do about climate change. What gives one crisis traction and the other none? Covid-19 is not a threat to the vast majority of humans, but the lockdowns affect everyone. Why aren't we this selfless regarding other issues like inequality and climate change?

The main difference is that governments think that they have to act immediately and comprehensively to respond to this threat. If climate change was going to happen in the next six months then they would act to the same degree.
Also there is the image of bodies on morgue tables, which doesn't play well with your electorate.

I think at least part of the answer lies in our myths and fears.

Yes, there is an innate fear of pandemic, whereas climate change is some distant idea for many people.

There is a consideration which I am sure was in the minds of the experts and politicians, at the point when more severe restrictions where contemplated. Which is that it would be impossible to avoid social and civil collapse in cities and then in the wider area if no action had been taken and there were many thousands of seriously ill people in tents and sports halls. So I think the measures are more to prevent this collapse than to avoid the death of a few million old and vulnerable people.

Such collapse would inevitably result in many more deaths than one would expect from the virus in isolation.

Out of 20,338 people tested in Britain for covid-19 164 people have the disease. The test itself is 97% accurate. You take the test and it comes back positive. What's the chance you actually have it based on this single test?

Apparently the answer is something like 21%? — Michael

Yes, it's 21%.

However the actual coronavirus tests will have a specificity of 100% if the same as the WHO tests:

The group verified the test in the absence of SARS-CoV-2 isolates or patient samples but confirmed its specificity against 297 clinical samples from patients with various other respiratory infections. This formed the basis of shipments of 250,000 kits, which the World Health Organization (WHO) dispatched to 159 laboratories across the globe in recent weeks. — Nature - Coronavirus and the race to distribute reliable diagnostics

Which references this paper:

The samples contained the broadest range of respiratory agents possible and reflected the general spectrum of virus concentrations encountered in diagnostic laboratories in these countries (Table 2). In total, this testing yielded no false positive outcomes. — Detection of 2019 novel coronavirus (2019-nCoV) by real-time RT-PCR

COVID and the looming debt crisis:

"According to the Institute of International Finance, total debt reached $253 trillion in late 2019, or the equivalent of 322% of global GDP – the highest it has ever been. Now that large parts of Europe and North America are following China in imposing far-reaching lockdowns, concerns are growing over the viability of this enormous debt pile. In the sharp economic contraction of the next few months, widely expected to become the worst in peacetime history, countless borrowers will struggle to repay their debts. This in turn risks unleashing a major international debt crisis that will make the market crash and global recession of 2008–’09 look like child’s play.

[In '08]...The world’s leading central banks soon joined this effort to preserve the financialised world economy, cutting interest rates to historic lows and pumping the equivalent of over $11 trillion worth of new money into circulation through their quantitative easing (QE) programmes. These dramatic monetary interventions helped stave off a total collapse of the global financial system, but they came at the cost of a fresh wave of speculative investment and a rapid increase in global debt levels, which has left the world economy extremely vulnerable to an unforeseen external shock.

And what a shock we got: a near-complete shutdown of productive and commercial activity in some of the world’s leading economies, combined with a collapse in the oil price followed by a synchronised and virtually instantaneous crash of money and capital markets, which threatened to freeze up international credit and payments systems, amidst concerns over collapsing global supply chains and skyrocketing unemployment levels. If ever there was a perfect storm, this has to be it.

...Unlike the new strain of coronavirus that triggered this medical crisis, however, the mountains of debt that now threaten to sink the global economy are no force of nature: they are man-made, and largely a consequence of the particular ways in which policymakers chose to deal with the last financial crisis."

https://www.tribunemag.co.uk/2020/03/the-coming-debt-deluge

Who is that debt owed to?

How do they know if a test is a false positive? Wouldn't they need some other, better, test to show that it's positive? Given that some people are asymptomatic, you can't look to symptoms as a measure.

Who is that debt owed to?

To each other.

What they could do is right down their debts and see how many noughts there are and then agree to cross off some of those noughts.

Problem solved.

Banks, governments, corporations, investors, individuals - part of the problem is that everyone owes each other and uses debt to pay for debt and because money has been so cheap (credit is - or was - so massively available) it's gotten easier and easier to think this bonanza can just keep going.

You can eradicate false positives by retesting. In the Netherlands each sample is tested twice by separate laboraties to lower the likelihood of false positives. In case of contrary outcomes, you test again.

No harm no foul for Senator Burr either. Just being optimistic with the public while privately telling his friends there was a disaster coming and selling hundreds of thousands of dollars in hotel and other stocks. Politicians like Trump and Burr j — Baden

I have no sympathy for this Burr guy, but how do get to Trump? Trump separated himself from his business when taking office, and his business IS friggin hotels, which is the sector most hard hit by this.

This TDS really screws with peoples heads.

There is always some time frame in which data fits an exponential growth curve! Or logarithmic. Or linear. Or better yet polynomial — SophistiCat

Yes! You finally get it. If we're interested in a time frame where exponential growth is accurate (or accurate enough for our purposes) it's perfectly valid to say the phenomenon has grown exponentially from A to B points in time.

Such as biologists saying bacteria grow exponentially when there is enough food.

What you don't seem to get is that your criticism of using the term "experiential growth" is the same for all curves (and continuous curves themselves don't "match" any phenomena because all measurement is fundamentally discrete).

If you say "oh, it's not exponential it's logistic" I can say "ahh, it's only logistic until it isn't, what if the virus doesn't behave in a logistic curve due to reinfection, or what if we wait long enough for new generations to be born that aren't immune and the virus re-emerges as a pandemic; see, haha, only logistic until it isn't!! hahaha".

What would your answer be? Logistic is a good curve to approximate infection growth on the time scale we're interested in, of this first wave of infection.

And, by the same logic "exponential" is a good curve to approximate the first phases of an outbreak, as other factors that reduce exponential growth are insignificant on the time scale of the initial outbreak.

Why are we interested in this time-scale? Because that's when the medical system is overwhelmed and governments decide on a policy response to stop the exponential growth that would occur if no action is taken.

Why is exponential a good description? Because the corrective terms and factors on this time scale would make no significant contribution to predicting the initial outbreak, so the equation can just be simplified to an exponential one, and it's useless pedantry to keep terms that simplify to zero.

- it can be made to fit any curve over any time scale. But no scientist in their right mind would propose an exponential growth model just because you can fit an exponential curve to two consecutive points. This is not how scientific modeling works. — SophistiCat

No where did I say we only have two data points, so I don't know where this strawman comes from, but it's not even a correct argument against the strawman you've created!

Scientists fit curves to small amounts of data, even 2 data points, all the time, and then debate which projection is justified based on either the trend in the data (as more data comes in) or proposed mechanisms.

Yes, no scientist would insist it's one projection rather than another without some argument, but fitting different curves to data and then debating what is "overfitting" and what isn't, what is the clear trend in different time frames, what mechanisms might change the trend in different time frames, is basically the second lesson of data analysis (after how to fit curves to data to begin with).

Which is exactly what I'm proposing!

Maybe growth will reduce close to zero soon in Italy, follow such a logistic projection (still not projected "forever" but only a good prediction insofar as quarantine measures are maintained, virus doesn't mutate to be more virulent to defeat quarantine; still only "logistic until it isn't" as you say) where new cases start to approach zero.

Maybe not! Maybe it will follow a projection where growth continues to be some significant percentage of the population, under the previous policy measures, and a exponential curve is a predicts accurately the future states over the time frame we're interested in (corrective terms to the logistic function are insignificant). Maybe the new policy measures, of marshal law, will get it under control ... maybe not!

Now, you can say "oh, well, I still don't like the word exponential here", but I'm just using the terminology epidemiologists and biologists use; you asked for a source, I provided you a citation of such language under the heading "exponential growth" in wikipedia. Why do scientists talk like this? Because the same kind of criticism can be brought in for everything, you could walk into any math class explaining exponential growth in terms of interest payments "gotcha! exponential interest growth is impossible because eventually no ledger could keep track of the numbers, so it's really an exponential added to a step function that reduces growth to zero when the entire accessible universe has been filled up tracking this number; why is this a foolish criticism? Because that time-frame is of no interest to the phenomena being discussed".

Why this is relevant (that we don't know yet what projection will actually be true; and we don't know which policy measure contributed exactly what amount in controlling the virus) is because it informs risk analysis.

A lot of people assumed Wuhan was a "worst case scenario" and that therefore Italy will follow a projection similar to Wuhan ("Italy is X weeks behind Wuhan" is a phrase that would pop up) essentially implying that Wuhan presented a upper-bound on "how bad it can be". However, Italy has now broken with the Wuhan pattern (of leveling off in growth about now). This could be due to the Wuhan scenario being mostly fraudulent numbers by the Chinese, or it could be that Italy didn't do a good enough policy response or, even with a similar policy response, conditions are more favourable to the virus in Italy or it has mutated to be more virulent (lack of adequate protection of health care workers creates the conditions of evolutionary pressure for the virus to become more virulent).

So Italy is now becoming the new worst case scenario.

Other places with an even later or even weaker, or both, response can make an even worse case scenario compared to Italy.

How do they know if a test is a false positive? Wouldn't they need some other, better, test to show that it's positive? — Michael

For specificity testing, you don't need a better comparative test because you use known negative samples that demonstrate that your test doesn't give false positives.

There are caveats on that, but that's the essence. You're looking at the gold standard test. The test will only be positive when the virus signature is detected.

Given that some people are asymptomatic, you can't look to symptoms as a measure. — Michael

Right.

You have to have quite a long memory for this "games" to be played out, because the actual death toll will likely be only known much later as people do extensive study just who died of the corona-virus. If an old person doesn't go to the hospital and dies at home, you think they will check all the people for corona-virus? Or was it the flu? Hence you have at first quite a wild range of numbers only for history later to correct it.

That happened with swineflu epidemic of 2009. Report from 2013, four years later:

The swine-flu pandemic of 2009 may have killed up to 203,000 people worldwide—10 times higher than the first estimates based on the number of cases confirmed by lab tests, according to a new analysis by an international group of scientists.

The researchers also found almost 20-fold higher rates of respiratory deaths in some countries in the Americas than in Europe. Looking only at deaths from pneumonia that may have been caused by the flu, they found that Mexico, Argentina and Brazil had the highest death rates from the pandemic in the world. The toll was far lower in New Zealand, Australia and most parts of Europe, according to the study, published today (Nov. 26) in the journal PLOS Medicine.

The new estimates are in line with a previous study published last year that used a different statistical strategy to evaluate the impact of the pandemic caused by the H1N1 virus. However, that study, which was done before countries' data on overall death rates in 2009 had become available, found that the majority of deaths occurred in Africa and Southeast Asia.

The new analysis "confirms that the H1N1 virus killed many more people globally than originally believed," said study author Lone Simonsen, a research professor at The George Washington University in Washington, D.C. "We also found that the mortality burden of this pandemic fell most heavily on younger people and those living in certain parts of the Americas."

Maybe an exercise in Bayesian statistics? I.e., a concern with false positives? — tim wood

Out of 20,338 people tested in Britain for covid-19 164 people have the disease. The test itself is 97% accurate. You take the test and it comes back positive. What's the chance you actually have it based on this single test?

Apparently the answer is something like 21%? — Michael

Good point. 97% accuracy isn't great. — Baden

These observations are correct on an individual level.

The reason for mass testing is systemic. Even ignoring how to increase the accuracy of the tests (multiple tests etc. which could even be counter productive, see below); having more positives than what's expected from false-positives is going to inform that community transmission is happening in the region, first of all and some policy response is needed (obviously, we'd want to save tests until community spread, and so it's more random sampling than mass testing in this "mode").

Second, some of the positives are going to be true-positives, but even if these are 20% of all positives (true and false positives), you can still just quarantine all positives anyways. Since the course of action of asymptomatics is to just self-isolate until there are symptoms, there's not really much cost to the false-positive (to society).

Furthermore, if society is going into lockdown anyways, then asymptomatic false positives just end up with "more motivation" to lockdown. So it actually makes the situation marginally better with some percentage trying to "super lockdown", thinking they have the virus even if they don't (mind trip, both not harmful to society).

Of course, false-negatives create the exact opposite behaviour, so the above benefits only arise if the chance of false-negatives are significantly lower than false-positives.

So, mass testing can still have a systemic policy effect even if results of positive tests are not used as a basis of diagnosis; diagnosis could be based on symptoms if and when they develop.

By diagnosing based on symptoms rather than serial testing the positives to identify the fasle-positives, this makes more tests available for the mass-testing policy. More mass testing without diagnosis can be more effective use of tests than trying to diagnose (to separate true-positives from false-positives through more testing, which if chances were only 20% for the first test a second test may still be inadequate to be highly confident).

In otherwords, there's the Bayesian statistics for the case of the individual, but also Bayesian statistics for policy. Mass testing will give better information on the sate of the outbreak than no mass testing, even if each individual positive is only marginally more informed. Isolative measures of all positives, both true and false, is going to significantly slow the spread of the virus even without diagnosing anyone until they have symptoms, as well as inform and track the effectiveness of policy measures sooner than the ultimate evidence of deaths.

Indeed, considering all this, the argument can be made that testing symptomatics "who want to know" is less informative (as the outbreak progresses there's less and less chance it's going to be something else) than mass testing to find asymptomatics in a large net that catches both true and false positives.