Nate Silver

Nate Silver @NateSilver538

Unless these tests are *very* accurate—and they probably won't be, since it's still the early stages of this crisis—this renders antibody studies not terribly useful in places where the underlying incidence of a disease in a population is low (say, under 5%). — PolitiTweet.org

Nate Silver @NateSilver538

Technically, the issue is not the false positive rate per se but that *we don't know what the true false positive rate* is. If we knew false positives were *exactly* 1%, we could correct for it. But it could be 0.5% or 2.5% or who knows—maybe not what the manufacturer claims. — PolitiTweet.org

Nate Silver @NateSilver538

Intuitively, the false positive problem should be easy to see. If a test produces, say, 1-2% false positives among healthy people, then results saying that 2-3% of people tested positive in a given region just won't really tell you very much. — PolitiTweet.org

Nate Silver @NateSilver538

Test accuracy has gotten the most discussion here. These tests aren't perfect. They produce false positives (it says you had it when you didn't) and false negatives (it says you didn't when you did). — PolitiTweet.org

Nate Silver @NateSilver538

OK, quick thread on the 3 major types of statistical issues in seroprevalence studies—that is, studies that try to figure out how many people had COVID-19, like the one released in NY yesterday. These are: 1) Test accuracy (false negatives/positives) 2) Sample selection 3) Lags — PolitiTweet.org

Nate Silver @NateSilver538

"Was that flu I had in January the coronavirus?" is like a textbook case of whether you understand Bayes' theorem. https://t.co/g0ZzM425ue — PolitiTweet.org

Nate Silver @NateSilver538

BUT: A lot of things chip away at the margins. Technology and new drugs/therapies help. We never get "enough" testing or contact tracing but we do get *some*. We learn more about the disease. We make improvements through trial-and-error. — PolitiTweet.org

Nate Silver @NateSilver538

In short (my version not Tyler's): it's a slog; there's never really any plan; we never really decide what trade-offs we're willing to make and so we careen back and forth and to some extent get the worst of both worlds. (The next tweet is more optimistic.) — PolitiTweet.org

Nate Silver @NateSilver538

I was trying to think through "There are so many unknowns, but if you had just one modal prediction for what the next 18 months look like, what would it say?" and I think it would probably would look a lot like @tylercowen's from a couple weeks ago: https://t.co/2xJlkTC0CM https://t.co/1dCnYjDYw8 — PolitiTweet.org

Nate Silver @NateSilver538

@JADubin5 Actually it's SARS-CoV-2 — PolitiTweet.org

Nate Silver @NateSilver538

@chrislhayes "This wave" is an important stipulation. That's not meant to sound super-duper pessimistic. There are rays of light. But, I think I have less idea than ever what our plan is to deal with this over the next XX months. — PolitiTweet.org

Nate Silver @NateSilver538

With that said, this is really not a national epidemic. It's a lot of state and local ones. And while some of the most problematic places from a few weeks ago are clearly getting better, many other places are struggling to bend the curve. https://t.co/7CxfmiXPWZ — PolitiTweet.org

Nate Silver @NateSilver538

One of the better-seeming days in a while. Deaths down from a week earlier. Tests up, and it wasn't just a one-day quirk like yesterday's dump of California numbers. — PolitiTweet.org

Nate Silver @NateSilver538

US daily numbers via @COVID19Tracking: Newly-reported deaths: Today: 1,911 Yesterday: 2,108 One week ago (4/17): 2,136 Newly-reported cases: T: 32K Y: 27K 4/17: 31K Newly-reported tests: T: 193K Y: 311K 4/17: 158K Share of positive tests: T: 17% Y: 9% 4/17: 19% — PolitiTweet.org

Nate Silver @NateSilver538

RT @trvrb: @NateSilver538 It’s possible there are strain differences in severity, but we don’t have evidence of this at this point. I want:… — PolitiTweet.org

Nate Silver @NateSilver538

I'm overdue to tweet out a couple of good podcasts I've been on! Here's one with @ASlavitt where we went very deep on election data, coronavirus data, and all the places in between. https://t.co/lVzmEtzvIk — PolitiTweet.org

Nate Silver @NateSilver538

@trvrb Do you think IFR is likely to vary meaningfully based on the strain or is largely a lot of unfounded speculation? — PolitiTweet.org

Nate Silver @NateSilver538

@CarlBialik Nope, not on its own, given how fast COVID spreads. However, if you're combining with social distancing, it can have some effect and tip R further below 1. NYC's curve has been bending down faster than other places (some not bending down at all) and that may be part of why IMO. — PolitiTweet.org

Nate Silver @NateSilver538

@CarlBialik That's because it can have meaningful effects at slowing things down even before you hit the threshold, but also there's momentum/overshoot, so some people will still get infected even after you hit the threshold. — PolitiTweet.org

Nate Silver @NateSilver538

@CarlBialik "Herd immunity" is another one of those terms that gets used in a few different ways. I tend *not* to like the usage where it's presumed to refer to some strict threshold. — PolitiTweet.org

Nate Silver @NateSilver538

The US really, really needs to work on its tools in addition to social distancing—meaning: testing, contact tracing, better protocols for handling sick individuals, etc. That's our way out of this in the medium term. We need the whole arsenal. — PolitiTweet.org

Nate Silver @NateSilver538

Anyway, in most parts of the country, it's premature to have the re-opening conversation. Either cases are still steady/increasing, or (as in NY) they're declining but still high in absolute numbers. There are perhaps some exceptions West of the Mississippi. — PolitiTweet.org

Nate Silver @NateSilver538

Essential workers are another challenge. The burden falls heavily on them. And it may make it hard to slow the spread too much in the population as a whole; it just concentrates cases in certain mostly disadvantaged subgroups. https://t.co/Y1nLtzKLzQ — PolitiTweet.org

Nate Silver @NateSilver538

Also, some of the success stories have factors that aren't easily replicated elsewhere. If ~20-25% of the NYC population had COVID-19, then herd immunity will knock down R somewhat, but that's not how you'd want to knock down R. Hot weather may have helped in Louisiana. — PolitiTweet.org

Nate Silver @NateSilver538

There definitely *are* success stories, foremost among them New York, Louisiana and Michigan. But in the average US state, it's not clear whether we're currently getting R<1. — PolitiTweet.org

Nate Silver @NateSilver538

I *am* trying to account for testing volume here, which is a lot of work. If you look at the estimated share of positive tests, it's still increasing in many places, or at least it was as of a few days ago. https://t.co/zIntxFJGXO — PolitiTweet.org

Nate Silver @NateSilver538

The US as a whole hasn't really turned the corner yet. Coronavirus cases are still at or near their peak in about half of US states. https://t.co/7CxfmiXPWZ — PolitiTweet.org

Nate Silver @NateSilver538

In a place where the underlying incidence of disease spread is low (Santa Clara), false positives are often the most pertinent bias and will tend to bias the numbers upward. The underlying incidence is *not* low in the NYC metro area, however. — PolitiTweet.org

Nate Silver @NateSilver538

To repeat: In these studies, there are many potential biases that can run in both directions. I think people mis-learned lessons from e.g. the not-very-well-done Santa Clara Co. study, and therefore assume the biases inherently tend to be upward. — PolitiTweet.org

Nate Silver @NateSilver538

I don't think that's at all clear. My initial reaction was the opposite. People out-and-about are less likely to have been feeling symptoms so that could bias the numbers downward. https://t.co/RHkJlXpSht — PolitiTweet.org