Nate Silver

Nate Silver @NateSilver538

Yeah. I think it's a bit irresponsible for the media to lean into the ZOMG POPULIST REVOLT! angle until the polling changes. Which it might later on, but has only done so at the margin so far. Or there's other evidence this is more widespread. — PolitiTweet.org

Nate Silver @NateSilver538

RT @FiveThirtyEight: Infectious-disease experts think the U.S. COVID-19 death toll will hit 50,000 by the end of April. https://t.co/Ddg4xi… — PolitiTweet.org

Nate Silver @NateSilver538

The federal response has been poor, to say the least, but states and metro areas joining together in deciding on how to re-open stuff seems wise and something that could equally well be happening under a good federal response. — PolitiTweet.org

Nate Silver @NateSilver538

RT @micahcohen: 😬 https://t.co/gafQHzeep1 https://t.co/ChxupNZ5fb — PolitiTweet.org

Nate Silver @NateSilver538

This can be defined mathematically: you want to design a model as to minimize serial correlation. Some experts prefer forecasts that change only gradually so they won't get criticized for changing their mind. But this intuition is often wrong. More here: https://t.co/0vo0wU4sXa — PolitiTweet.org

Nate Silver @NateSilver538

But unless you've built a model that *updates in real time*, I think it's easy to *underestimate* how much a well-designed model *should* change—certainly not all the time, but in response to high-leverage data. Most experts' intuitions are a bit miscalibrated here. — PolitiTweet.org

Nate Silver @NateSilver538

My default answer is "sparingly; you should avoid it". But I'm working in fields (i.e. electoral politics) where the structures are well-established going in. I am perhaps more sympathetic to structural changes with something like COVID-19 where we're all learning in real time. — PolitiTweet.org

Nate Silver @NateSilver538

Another question is: should the structure/parameters/assumptions of the model change as you collect more data? In other words, not just the inputs, but the way you're processing those inputs? And maybe which inputs you're looking at? That's complicated. — PolitiTweet.org

Nate Silver @NateSilver538

The death projections in their US model have shifted around within a range of ~60K to ~90K. That strikes me as quite reasonable, under the circumstances. Other things have changed more, i.e. their projections on hospitalizations have decreased by 3x to 5x. That's not great. — PolitiTweet.org

Nate Silver @NateSilver538

There are some good critiques of the IHME model in here IMO but this line bothered me. Model outputs *should* change in response to new data and when you're dealing with nonlinear systems, small changes can have a reasonably large effect. https://t.co/rWd9SgFp6y https://t.co/wQRLXLLzx2 — PolitiTweet.org

Nate Silver @NateSilver538

RT @onyxfish: Missing the NBA? Why not pass the time watching some of the best games ever? An epic new interactive from @ryanabest and @Nei… — PolitiTweet.org

Nate Silver @NateSilver538

@joshtpm I think people started to figure it out before the city government did. Almost day to day, you could see things thinning out by 20% at a time. — PolitiTweet.org

Nate Silver @NateSilver538

The days when they were like "nothing to worry about, but maybe you should avoid the subways!" were the most worrying days of all. — PolitiTweet.org

Nate Silver @NateSilver538

OK, maybe this is a poorly-framed question. But I have learned to be skeptical of "clinical trial shows promising results!" headlines. What would experts say the benchmarks are for something they'd actually consider to be a development worth getting excited about? — PolitiTweet.org

Nate Silver @NateSilver538

But not on No. 2. There have been days when there were fewer deaths than the previous day. But these may reflect day-of-week effects. Not yet any days when there were fewer deaths than 7 days ago. Once that happens a few times, that's when we can start asking if there's a peak. — PolitiTweet.org

Nate Silver @NateSilver538

The milestones I'm mostly looking at are: 1) Are there fewer cases (and/or a lower share of positives) than the same day a week ago? 2) Are there fewer deaths than the same day a week ago? For question No. 1, the answer is increasingly *yes*. Which is good news. — PolitiTweet.org

Nate Silver @NateSilver538

Daily US numbers via @COVID19Tracking: Newly-reported deaths: Today: 2,136 Yesterday: 2,492 One Week Ago (4/9): 1,877 Newly-reported cases: T: 31K Y: 30K 4/9: 34K Newly-reported tests: T: 158K Y: 161K 4/9: 163K Share of tests positive: T: 19% Y: 19% 4/9: 21% — PolitiTweet.org

Nate Silver @NateSilver538

RT @MysteryPollster: @SeanTrende @NateSilver538 Just published this, but thought it would be of interest & relevant to this thread… https:/… — PolitiTweet.org

Nate Silver @NateSilver538

@PatrickRuffini I'm not the person to ask. My understanding is something like "to a first approximation, likely yes, at least for now, but there are uncertainties and complications, especially re: how long immunity lasts and how complete it is." https://t.co/VKIyTMK2Eu — PolitiTweet.org

Nate Silver @NateSilver538

But this highlights how we really, really, need more tools than just social distancing, i.e. testing, tracing, screening, and better protocols around infected cases. If you're relying on social distancing alone, you seem to need a lot of it to keep R<1. — PolitiTweet.org

Nate Silver @NateSilver538

Other factors that seem relevant here: 1) How much herd immunity is there in the very worst affected areas, e.g. Lombardy and NYC? 2) What is the R among different subgroups, i.e. health care workers, other essential workers, rest-of-population? — PolitiTweet.org

Nate Silver @NateSilver538

Per my back-of-envelope, possibly a *bit* lower than that if you look at the share of positive tests in Italy. But, yeah, maybe not a lot lower, meaning there may not be a ton of margin for error in re-opening stuff unless other factors reduce the spread. Same issue in NYC. — PolitiTweet.org

Nate Silver @NateSilver538

In this story, there's lots of reporting on Sanders-friendly groups and people who *do* want Warren to be chosen, but the story is framed the opposite way, perhaps because reporters overstate the influence of the Very Online crowd (+drama makes headlines). https://t.co/4SJ73zuJg1 — PolitiTweet.org

Nate Silver @NateSilver538

Relatedly, the Very Online folks don't have much leverage over Biden because it's not clear what would make them happy, and the things that *might* actually make them happy (but probably not) are things that Biden would likely never consider. — PolitiTweet.org

Nate Silver @NateSilver538

Warren being chosen as VP would be a good example of something that normal Sanders supporters would like, and the fact that a few Very Online ones might not like it is a sign that they aren't really speaking for the majority. — PolitiTweet.org

Nate Silver @NateSilver538

Keep in mind that, at some points during the worst of the upslope in the US and Europe, new cases on a given day could equal 20% or even 30% of total cases. So being at 10% or 15% isn't *that* fast. Still, these are places to monitor carefully. — PolitiTweet.org

Nate Silver @NateSilver538

The list: Russia Belarus Ukraine Uzbekistan Brazil Argentina Peru Singapore Bangladesh https://t.co/9fFIJJjqhz — PolitiTweet.org

Nate Silver @NateSilver538

This is not a perfect measure, but here is a list of countries where newly-diagnosed cases yesterday were at least 10% of total cases, a possible indication of places where things are getting worse. Minimum of 1000 total cases to make this list. — PolitiTweet.org

Nate Silver @NateSilver538

To a first approximation, under-reporting is likely to be worse in areas where the epidemic is worse. In these places people may avoid, or may be turned away from, getting care in hospitals. It's the at-home (and often nursing home) deaths that are being missed. — PolitiTweet.org

Nate Silver @NateSilver538

There's a certain group of people who correctly note that COVID-19 deaths are underreported but then assume without evidence that they're underreported in a way that just so happens to fit their priors. — PolitiTweet.org