Nate Silver

Nate Silver @NateSilver538

@jeffhauser This data is from hospitals not just positive tests, I believe. But that might worsen the biases you describe, i.e. the older family member comes to the hospital because they have worse symptoms than the younger working family member. — PolitiTweet.org

Nate Silver @NateSilver538

@asymmetricinfo @interfluidity Yeah, I agree it's all very complicated. And none of the models seem to deal super well with what happens when people have been in lockdown and those lockdowns are relaxed, which could have further unpredictable effects. — PolitiTweet.org

Nate Silver @NateSilver538

@asymmetricinfo @interfluidity I think the broad intuition is also incorrect. Most models suggest heterogeneity *lowers* the nationwide herd immunity threshold, although there are fierce debates about whether this is a meaningful effect or a negligible one. https://t.co/VdTPR0CJ2C https://t.co/FzHLB4yLvx — PolitiTweet.org

Nate Silver @NateSilver538

@asymmetricinfo @interfluidity This isn't correct because R varies by subgroup (the estimate of 2.5 or whatever is a population average) and therefore the herd immunity threshold varies by subgroup. — PolitiTweet.org

Nate Silver @NateSilver538

*Either* the worker *or* the vulnerable household member could use the hotel room, whatever works best for the family. — PolitiTweet.org

Nate Silver @NateSilver538

Yeah. Like why not open up unused hotel space for essential workers (especially in health care) who live with older or immunocompromised people in their household? — PolitiTweet.org

Nate Silver @NateSilver538

Why be careful? One essential worker could spread to several family members, especially in close/impoverished living quarters. — PolitiTweet.org

Nate Silver @NateSilver538

New data from New York state suggests that new cases are not primary occurring among people working in essential services but among older people at home. Although, one might need to be careful with this. https://t.co/ea152QkONJ — PolitiTweet.org

Nate Silver @NateSilver538

This is why I'm urging caution about what to expect in the data from states that "open up". It's not that I think opening up is a good idea—I think most states should wait. It's because the scope is narrow enough that it MAY get swamped by other factors because the data is noisy. — PolitiTweet.org

Nate Silver @NateSilver538

Maybe we go from a 25 to 35 or 40 in the "opening up" states...and in the "closed" states, maybe we go from 25 to 30 or 35 anyway because per a variety of data, people are getting fatigued with social distancing and it's hard to stop people from e.g. going to a friend's place. — PolitiTweet.org

Nate Silver @NateSilver538

So if 0 is a strict Wuhan or Italy-style lockdown and 100 is "oblivious pre-COVID life", we are not going from 0 to 100. We were never *at* 0, for one thing; many activities were still allowed under lockdown. So let's say we were at a 25. — PolitiTweet.org

Nate Silver @NateSilver538

Moreover, within the category of things that are opening up, both polls and initial reporting suggest that many people will stay at home & decline to participate in them, and many of the people that do participate will take precautions, as will most of the businesses themselves. — PolitiTweet.org

Nate Silver @NateSilver538

i.e. many states have plans such as: restaurants and personal care businesses can reopen, but schools stay closed and most businesses continue to encourage (or require) work from home. Parks may have been open all along & bans on small gatherings were hard to enforce all along. — PolitiTweet.org

Nate Silver @NateSilver538

You could maybe divide the remaining activities into two groups: "work + school" and "discretionary, small-scale social activity" (e.g. restaurants, parks, personal care, small household gatherings). Most "opening up" involves only some fraction of these activities resuming... — PolitiTweet.org

Nate Silver @NateSilver538

On the one hand, some (perhaps much) transmission occurs within households, or among people performing essential services. Those things never closed. On the other hand, some transmission occurs at mass gatherings and other superspreading events. Those aren't opening up (yet). — PolitiTweet.org

Nate Silver @NateSilver538

Part of the disconnect on "opening up" is because some people assume it means "back to normal"—when in fact the activities that fall under "opening up" (at least the ones states are considering and people are arguing about right now) are often in relatively narrow categories. — PolitiTweet.org

Nate Silver @NateSilver538

RT @mlipsitch: Clinical epidemiologists who are rational about decision making under uncertainty understand that with noncommunicable disea… — PolitiTweet.org

Nate Silver @NateSilver538

@mattyglesias Yeah. A lot to be worried about. — PolitiTweet.org

Nate Silver @NateSilver538

DC/MD/VA is one of the more problematic areas. It's dense, but hasn't had super strict social distancing, and cases are probably increasing (albeit modestly) even once you account for testing volume. — PolitiTweet.org

Nate Silver @NateSilver538

Super interesting stuff here — PolitiTweet.org

Nate Silver @NateSilver538

It's not a question of innumeracy because the math isn't complicated. It's more in the category of "too good to check", with not bothering to be rigorous when the surface (but perhaps misleading upon even mild scrutiny) interpretation of the data meets a certain narrative. — PolitiTweet.org

Nate Silver @NateSilver538

Another mistake is not understanding lags in the data. If someone gets infected, it will likely take ~2 weeks before they get symptoms, get tested, and the test shows up in the data. Yet many articles assume that cases will *instantly* rise when regulations are changed. — PolitiTweet.org

Nate Silver @NateSilver538

One basic mistake is in not accounting for tests. If a state doubles its testing it is going to see far more cases. Many articles about big spikes don't even bother noting when they were caused by tests increasing. The math isn't hard. It's a basic failure to provide context. — PolitiTweet.org

Nate Silver @NateSilver538

I dunno, this feels like an excuse for the media to not even bother to aim for rigorousness when evaluating coronavirus numbers. And there are some fairly basic mistakes being made. — PolitiTweet.org

Nate Silver @NateSilver538

I've seen many times in election coverage when the media narrative gets overextended based on dubious analysis of noisy data and then abruptly has to reverse itself and winds up overcompensating in the other direction. — PolitiTweet.org

Nate Silver @NateSilver538

Meanwhile, the media is touting what are supposedly extremely dire projections by the White House (200K cases/day) but which actually have an extremely unclear provenance, and which could be used to make merely-awful numbers look good by comparison. — PolitiTweet.org

Nate Silver @NateSilver538

I could be wrong. But the lessons if you've looked at the different states and countries is that the data is noisy and fairly unpredictable, and often only loosely matches media narratives about high-profile events (e.g. Florida beaches) and who's generally doing a good job. — PolitiTweet.org

Nate Silver @NateSilver538

Moreover, the correlation between changes in state policies and changes in the # of cases is liable to be noisy. There may/probably will be something there if you look rigorously. But for every case that fit the prevailing narrative, there may be others that contradict it. — PolitiTweet.org

Nate Silver @NateSilver538

While it is possible that we will see some big spikes as a result of states re-opening, the base case is probably a continued mix of plateaus, mild declines and mild increases, perhaps in a slightly different/worse mix than before (i.e. some declines turn into plateaus). — PolitiTweet.org

Nate Silver @NateSilver538

Having done this for a while now, I would say two days is too soon to jump to any conclusions. If you see something for four or five days in a row, or say 7 days out of 8, then maybe we can start revising our priors. — PolitiTweet.org