outbreak! (ebola, sars, coronavirus, etc)

does this mean that generally speaking the more prevalent the illness, the higher the PPV, even if the specificity/sensitivity is relatively high?

yes

you can fill out this table with some simple round numbers and see how changing the prevalence, sens/spec, etc, one at a time changes the PPVs and NPVs

whoops https://www.medcalc.org/calc/diagnostic_test.php

So that would just have to shake out over time the more prevalent covid becomes, right? That is, the PPV is lower than 50% because it's not that widespread yet, so statistically, on paper, it's likely that a lot of those positives are false positives. But that's based on projection, right? It's also possible, if less statistically likely, that in practice all of those positives are in fact true positives?

FL has once again temporarily shutdown standalone bars, suspending alcohol service statewide (restaurants that serve food and alcohol can continue)

Meanwhile, Illinois continues to cautiously open up, encouraged by great numbers. The question in my mind is not will the states/places that half-assed it at best run into problems - duh - it's will the states/places that took things seriously and got good results be able to sustain those results as things open up again.

Florida supposedly had "good numbers" before re-opening, which is usually easy when you game the data like Desantis did/is doing.

speaking of,

"Florida is reporting 8,942 new covid-19 cases on Friday, blowing past its single-day high of 5,511 set on Wednesday. It is the 19th day in a row the state has hit a new average high. Average cases are now up about 77 percent from a week ago, and 526 percent since Memorial Day."

fuck.

Yeah, it's hard to hide that kind of data.

Rep. Mark Green, R-Tenn., is furious because, earlier today, he saw a jogger wearing a face mask. pic.twitter.com/OVyvP1v41k

— Alexander Nazaryan (@alexnazaryan) June 26, 2020

"Don't look back, you can never look back"

I've given up with Florida. we have a Governor who doesn't give a shit, we have a populace who doesn't give a shit.

you should maybe get out of there?

if I didn't have folks living here that might need me for emergencies, that ship woulda sailed a while ago.

course, all of my usual destinations are places that also aren't doing well with COVID, so it might be "midwestern state with 50 miles between houses"

will the states/places that took things seriously and got good results be able to sustain those results as things open up again.

I think the widely agreed answer among epidemiologists is 'no'. An increase in cases is unavoidable, along with an increase in hospitalizations and deaths. The opening up of businesses and small social gatherings is a solution to a different problem than "how do we keep the virus from spreading"; instead it is in answer to the profound social costs incurred by the indefinite extension of the shutdown, not just in terms of the economy and plummeting tax revenues, but also the massive mental stress on the whole population and the resulting social unrest and desperation that was emerging.

Some form of reopening was always bound to happen. What's most disheartening is that the time that should have been spent strengthening the public health infrastructure and preparing the nation to unify in the face of a dire necessity, was spent dithering, wallowing in incompetence, then sowing discord in search of political (and financial) gain. Now we just have chaos and confusion, and the people who are saying the right things about maintaining caution, distancing and wearing masks are drowned out or talked over by idiots and nonsense.

https://www.cnn.com/2020/06/26/politics/maskwearing-coronavirus-analysis/index.html

I've given up with Florida. we have a Governor who doesn't give a shit, we have a populace who doesn't give a shit.

― I hear that sometimes Satan wants to defund police (Neanderthal), Friday, June 26, 2020 10:37 AM (seventeen minutes ago) bookmarkflaglink

cmon over to arizona...

Also, come on, can we not to do the whole "just move" thing? It's rarely that easy.

not to mention moving from a place of high incidence to a place of low incidence is potentially a really good way to spread a disease

tbf it WOULD be easy for me compared to other people (if not for the aforementioned folks) in that I have an assload of savings, live alone, and a job that I can do in any state.

but yeah, I would definitely not do so unless I had a clean test first and self-quarantined for 14 days upon arrival.

truth be told though I'm thinking of moving into the ocean

No Covid in the ocean, afaict.

I agree that the "why don't you just move" schtick is always a lame reply but i would maybe suggest an exception for florida

there's also the "everybody I have ever known in my life sans about four people live in Florida" dilemma, but maybe we can all move and form a commune somewhere.

i should note that i lived in sarasota for nine months at one point, speaking from some experience

No Covid in the ocean, afaict.

― Josh in Chicago, Friday, June 26, 2020 2:08 PM (seven minutes ago)

very high mask use iirc

wait til the first person tries to fuck a whale

the arizona of the ocean

https://i.imgur.com/sHDCpfE.jpg

wtf are you all doing! where a mask! fuck

except all of the fish are like "i just want other fish to follow the law. and maybe the people telling us to wear masks should be declared domestic terrorists, is what many people are saying"

*fish stare blankly, almost like they don't need to blink at all, almost like they're not human*

"i just want to spend a weekend on the beach with my family!!!!!"

"i don't know whether to wear a mask! i need a world of probabilities between 0.1% and 99.9% to deliver a probability of 100% so that i have no uncertainty! for now i won't wear a mask tho"

that is my impersonation of coronavirus fish in arizona, that's all i got. i will go back to work now

lol thank you for that KM, sincerely

i made them arizona fish instead of florida fish, out of respect 2 u

https://www.youtube.com/watch?v=to1XKxFlb2I

You can see the actual moment when VP Pence's soul would've left his body if he still had a soul. pic.twitter.com/gmeVAUxTvf

— BrooklynDad_Defiant! (@mmpadellan) June 26, 2020

BREAKING: Tennessee sets an all-time daily high in new #COVID19 cases with 1,410, already making this the worst week of the pandemic 1/ https://t.co/54BWulnpYZ

— Phil Williams (@NC5PhilWilliams) June 26, 2020

what a fucking disaster! Who would have thought?

the great thing is that each state has different rules about opening and shutting down, and some are in coordination with each other, but most of them aren't! that way, the declining conditions in one state can be sure to affect the states around them as well, including the ones that were starting to do better, on an endless cycle! some have talked about a mysterious "united" entity that would have oversight of the whole thing, like a central effort to coordinate all of the different states into one single "plan". but how many people would that involve? and how can we see them all on the same screen at the same time? technically we could - sure look at the phantom menace trade federation scenes for a look at what that kind of giant central coordinated meeting of "united states" might look like - but there's a reason phantom menace is science-fiction, or fantasy and not a documentary. so for now we all just have to come up with our own plans and pray

I can cross three or four state lines in a 20-minute drive.

you are the new quad-state captain

I can cross three or four state lines in a 20-minute drive.

nice

Yes. Phase three should be "we're doing everything outside for now." https://t.co/W2ltjcFAKE

— 'Weird Alex' Pareene (@pareene) June 26, 2020

Thank you! That helps a lot. A follow up: those numbers seem to hinge mathematically/scientifically on a survey of the broader, general population where the vast majority of people have likely not had or even been exposed to Covid. But if you just tested people who had or suspected they had Covid, a sort of self-selecting (if not statistically sound, scientifically speaking) pool, wouldn't the PPV be higher? Maybe it isn't useful in terms of applicable data that can then be extrapolated across populations, but it seems that it could be useful for that specific demographic.

Right. A test with a high FPR becomes less and less useful if population prevalence goes down. Conversely, you can get useful information from a high FPR test if you restrict its use to populations who are more likely to have the condition. "Useful information" here means "you can maximize the number of people who really do have the condition".

The PPV is probably better than 50% in practice because the prevalence of antibodies *among people taking the test* is higher than the background population prevalence (you only take the test right now if you have reason to think you have had covid). But the corollary of that is: screening everyone for antibodies is not useful.

That is, you don't screen young people for diseases associated with old age, but if you limited screenings *exclusively* to old people, then would the PPV for Old Age Illness X be higher? Does it/can it work that way?

Exactly. The PPV of screening 100% of the population for breast cancer every year is very low because most people don't have breast cancer. But if you only screen women then PPV goes up. And if you only screen women over X it goes up more.

So in the case of the strep rapid test, it's hitting 98-99% specificity, which is good, but does't seem *that* much better than the 95% specificity of the Covid antibody test, and the sensitivity of both tests seem relatively similar (around 90%). Which may be why I'd heard that the Covid test, like strep, is more likely to yield false negatives than false positives, even if the false positive rate is still statistically something to be aware of.

Well, it depends on your perspective. If you're a health official, you're thinking in terms of testing a bunch of people. In that situation, the covid antibody test is more likely to yield false positives (in the sense that it's a more common outcome than a false negative) simply because most people don't have the antibody, so they logically can't have a false negative! There's just fewer opportuntities for the test to screw up in the sense of missing a case. There are lots of opportunities for it to screw up by seeing antibodies where they don't exist.

If you're a individual person deciding whether to have a test it's different. Absent other information, there's a 5% chance you have the antibodies. Before you take the test, there are two ways the test can go wrong for you: FP and FN. The probability of a FP is 95% (you're negative) * 5% (the test screws up) = 4.7%. The probability of a FN is 5% (you're positive) * 10% (the test screws up) = 0.5%. So yes, the probability of a false negative for a particular person who hasn't yet taken the test is higher. But once you take the test you can rule one of these two out so the relative probability of these two outcomes isn't super important.

One last point: it might not seem it, but 99% specificity is *a lot* better than 95% specificity. Think in terms of the FPR. That's 1% vs 5%. It's five times better! (This is why I like FPR and FNR btw, rather than specificity and sensitivity.)

You know how you hear that machine learning has gotten really good recently? It's a bunch of tasks we used to get right 95-97% of the time. Now it's 99%. That is a *HUGE* improvement and has qualitatively changed how useful ML is, even though 99 doesn't seem that much bigger than 95.

* "Useful information" here means "you can maximize the number of people who really do have the condition that you find".

Google searches related to loss of smell, past seven days https://t.co/rCeGQJmQo9 pic.twitter.com/aj8RaKkswW

— Christopher Ingraham (@_cingraham) June 26, 2020

oof

Dang

47,000+ new cases today

in the US