(untitled)

There's been a lot of talk around here about political aspects of the novel coronavirus, but almost no talk of quantitative approaches to modelling both the current state of coronavirus infection, R_0, risk of infection, stuff like that. A number that I'm particularly interested in the the probability that a random person in my county is infectious. To do this I look up the number of weekly new cases on the county public health website, multiply it by 4 (to account for a 4-week infectiousness period) and then multiply it by 1.2 to guess that 20% of coronavirus cases are never tested or have a false negative. Call that number N, my estimate for the number of currently infectious people in my county. Then I take the number of households in the county according to the census, since I believe that if someone shares a house with an infectious person they're probably infectious too. So I divide N by that number and get the probability that a random person in my county has the coronavirus. Right now it is about 0.013. When you factor in the fact that most people who test positive for coronavirus are either self-isolating or in the hospital, that's a very low chance that anyone I interact with is infectious! So I don't worry too much at all. How do you guys do this? Do you try to build forecast models? I tried to do that but the equation I got was nonlinear and chaotic and involved too many guessed parameters, so it was basically useless. What this thread is NOT for: pushing your view on virus politics, arguing for or against the existence of the virus, etc. This thread is only for quantitative analysis. Image not related.

I think what you really want to know is your own risk. For that you have to consider the distribution won't be uniform across your country because it will heavily rely on movement between cities. A city with air port / tourism / ... will do much worse than one of those pensioner cities where only old people live that never leave even their own home. Rich cities might do worse than poor cause more have money to vacation without giving a fuarrrk. Its as you say to many factors to consider this above astrology level. The conclusion i took from this is just stopped caring.

>>18976 I wrote county, not country. Counties are pretty small, sometimes smaller than cities.

>>18978 Admit read it wrong but its the same smaller. You got a hospital? Most will likely will be there. You got none? More distributed. Got a university? Many people came back from congresses with infections - ironically biologists researching it too, had one case in our department. There's clustering of patients either way. Your personal risk will depend on your % to interact with risk groups.

>>18980 indeed, these are all subjective rather than quantitative measures so I didn't include them in my OP.

If you don't go out, you don't get infected, simple as. NEETs against statistics.

>>19024 Food delivery, gg vault dweller

>>19025 I'm not a retarded consoomer who interacts with unwashed proles to get his overpriced msg pizza and supermarket dog food delivered by illegal immigrant who lives in unsanitary conditions with 15 other like him and their extended families being ideal for spread of such disease. I make all purchases twice a year in wholesale market at the nearest suburb location, only need to take train in off-peak hours, and bring a big trolley bag with me. Plus if I'd really have to order delivery, I can just quarantine all food for few weeks and blast it with ultraviolet lamps. Overall, this "pandemic" still much less of a problem for people under 70 who are not a braindead consumerist party animals with urge to go out in public places every single evening and smooch everyone there.

Won't work. You have to assume actual cases are at least four times larger than confirmed cases due to how some people exhibit minor symptoms.

The most experienced epidemic model programmers in the world were off by like 10-100x. As in we might as well have based our policy decisions on the divination of astrologers. So yeah, there's a lot of room for improvement with the models. I don't think a bunch of randos should be doing it, though. We all remember that asshole Tomas Pueyo—an armchair covid expert and tech entrepreneur at a remote learning startup, no conflict there holy ѕһіt. Maybe you could make a post-hoc model that better matches the arc of the virus, but that model would be taking into account our schizophrenic, piecemeal response strategies and essentially treating it as if it were a property of the virus itself. If you tried to compensate, you'd be introducing who knows how much inaccuracy with your wild assumptions. A monkey with a D-20 could do a better job than Neil Ferguson, well-known hack infamous for decades of fraudulent doomsaying. No fuсking idea why anybody listened to him. Probably because he was the only one unhinged enough to peg the numbers that high, so he got the most ratings. Who knows. But while anybody on this imageboard could do better than him, I don't think that it's possible to guess the range of values for enough of the variables (or indeed the existence of some necessary variables in the first place) before the spread. I'm not convinced that computer models are scientific in any configuration with something so badly understood, and it must be badly understood at the point in time when the model would be useful. I think we have this tendency to say computers and code equals science, but that is absolutely incorrect. This is just reading chicken entrails but with fancier and shinier guts made from silicon. It's still unscientific superstition nonetheless. https://www.spectator.co.uk/article/can-we-trust-covid-modelling-more-evidence-from-sweden https://swprs.org/a-swiss-doctor-on-covid-19/ One of the lines on these graphs is the result of modeling with different mitigation strategies of varying extreme restrictiveness. The other is the actual result in Sweden, the nation whose data the model was ran on, with nothing more than general public service announcements. These graphs should make anyone furious, or at the very least it should dissuade you from putting much stock in these preemptive computer models in the future. tl;dr: don't fiddle and pretend you're qualified to make models. That's how we got into this mess. The media listened to random computer nerds like us because they don't have any distinction between "guy at a remote learning service in Silicon Valley with an obvious bias" and "actual medical academic." Much of our public policy was lifted from Pueyo's fuсking Medium post. We should avoid this in the future if at all possible. I hope this is sufficiently on-topic and not full of politics, but I am also kind of baffled by your expectation that such a thing is currently possible to simply separate out like that. Obviously any computer model made at this point in time is going to have political bias and assumptions based on ideology; we're too close to have reliable data, so we have to choose which presentation to feed our models and that's political.

>>19038 And speaking of the improbability of subtracting politics from this subject: nobody can really decide if their content goes viral, as Pueyo's post did. Even if you somehow have the perfect non-partisanship of a saint, if your model shows an outcome supporting the agenda of a vague stochastic group of people on twitter or even an established corporate or political party that is pursuing a strategy organized at high levels, people will seize upon it and it may go viral and end up being the thing that wrecks the world economy for the next ten or twenty years... killing hundreds of millions globally with famines that were not even within the scope of your model. Whoops, oh well. Building a pandemic model is ambitious content, and even some yokel on an imageboard has a likelihood of getting international media attention if they achieved this at a sensitive time like back in March. You wouldn't necessarily have ideological control of your figures (and the massive errors that doubtlessly would be contained therein) once it took off. People would just stop listening to you and sneak their own commentary in next to cherry-picked results from your program. With information comes power, and you have a responsibility for unleashing that power. Don't be so sure this is just an experiment for fun; it could end up having real effects. It has happened before.

The easiest and most accurate way to gauge the true impact of coronavirus is to simply look up deaths from all causes this year and compare it to deaths from all causes in each of the previous 5 or so years. No politics, no testing errors, no statistics spinning, no propaganda, no bs. While the "official" deaths from coronavirus in the US are supposedly about 150k, the actual increase of deaths from all causes (over and above the average deaths from all causes) is about 220k. I'm not suggesting a conspiracy, just that there are plenty of good reasons why there's an undercount.

>>19040 Generally people want a separation of corona cases and cases caused by the measures against corona. You know all the freshly impoverished from business going rip who couldnt get treatment anymore for example.

>>19040 Nope. That's not easy or accurate or apolitical at all. You're totally missing the fact that there are other exceptional causes of death this year besides covid itself. The lockdowns cause enormous rates of excess death. There are counties in California which have recorded an entire year's worth of suicides in the past four weeks. People are avoiding hospitals and health care like the plague—literally—which has led to delayed procedures and diagnoses, increasing deaths in a massively significant manner. People are more sedentary and alcohol and substance use is through the roof. People are being evicted and encountering food insecurity at an incredibly alarming rate. Child neglect and accidents are extremely high, because there's no form of child care for families where both parents work full time and they choose to risk it because the only other option is not being able to house or feed themselves or their children. The thought doesn't even occur to you people that things can kill besides covid. It's so incredibly out of touch to not know that this is going on, or indeed to think that lockdowns would be fatality-free in the first place. I know many people affected by this already, and two suicides so far. I know five friends evicted out on the street and I've lost count of those who have lost their jobs. Some get unemployment to varying degrees, but the most poor and vulnerable were getting paid under the table... so nothing. Just instant homelessness and hunger. I've had covid myself, and known plenty of others who came down with it.. Verified by testing and including a person over 65. So far, nothing worse than a regular flu among these admittedly anecdotal experiences. All survived, no "long term damage" to be heard of. At least nothing that we can discern through the damage of trying to survive lockdown. Not everybody lives like the WFH and netflix crowd. Everyone's stressed and depressed and trying to eat cheap. By saying the excess deaths simply equal covid deaths, thus it is proven—to say that this is inherently accurate and apolitical—is itself an incredibly deceptive and political act. Even if you didn't realize it through ignorance. It's far more likely that covid is being overcounted, considering the methodology they are using to determine cause of death in these cases and the extreme age and number of comorbidities among the casualties. Covid hardly ever kills anyone who isn't already dying from at least one other thing, so naturally it's going to be difficult to work out the order of operations and ultimate culprit. But also: no, the majority of excess deaths are not undetected covid deaths. Lockdown is uniquely lethal. We'll be seeing more of this come to light in the next decade or so, kind of like war crimes slowly trickle out long after the war ends. The U.K. recently estimated two deaths from unnecessary restrictions for every three from covid. Like everything we're seeing this year, I expect the estimates for lockdown lethality to be underplayed and covid to be overplayed. It's all starting to slowly level out, and covid turns out to shrink and shrink the more data we have. What have we done? https://fee.org/articles/a-years-worth-of-suicide-attempts-in-four-weeks-the-unintended-consequences-of-covid-19-lockdowns/ https://www.telegraph.co.uk/news/2020/08/07/lockdown-killed-two-three-died-coronavirus/ https://www.ehstoday.com/health/article/21134240/a-pandemic-within-a-pandemic-substance-abuse-rises-amid-covid A reminder that in 1968, during a similar modern pandemic that had an even higher crude fatality rate than covid-19... Woodstock happened. And didn't result in mass deaths or shrill alarms about a "surge in cases" or whatnot. All of the lockdown stuff is not a standard response to any of this. It's entirely experimental, unprecedented, and experts have no data on how dangerous it is to lock up the world for half a year like this. It was not done in any kind of scientific or informed manner. The only data we have on this strategy is what has been collected this year. It was a blind leap. Experts didn't do this, fuсking Twitter did this. We are 100% going to see more deaths from lockdown policy than covid. It's just a question of whether it's two times or ten times or a hundred times. Yes, a hundred! That comes from the most deadly lockdown symptom of all: famine. Restriction of migratory agricultural workers is expected to result in catastrophic food insecurity for more than two hundred million people this year alone. And of course that will also effect collection of seed for the next harvest, so it's going to repeat next season. So that on its own is potentially about a hundred times the global death count of the virus. Even if we end up with a best-case scenario through pure luck, a rounding error on that figure would overshadow covid excess deaths. And this stuff takes a minute. So keep watching those damn stats next year, hopefully without a lockdown stubbornly persisting with no criteria even given for lifting it. https://news.un.org/en/story/2020/04/1062272 And of course, covid is hardly killing any children thank god. But Unicef is reporting up to 6000 child deaths are possible each day due to lockdown measures and complications: https://www.unicef.org/press-releases/covid-19-devastates-already-fragile-health-systems-over-6000-additional-children None of this would have been within the scope of expertise of medical experts. This is why we can never again allow this kind of autocratic policymaking to emerge from the advice of one extremely narrow field of research, much less a badly-designed computer simulation within that field. While it's a medical issue at its core, the world is an interconnected system which requires a hundred disciplines to begin to understand. Which is why we have governments and not just doctors, obviously. And that doesn't even get into the philosophical question of whether we should allow an academic to have total power to override constitutions and enact global martial law—even if they turn out to be 100% correct. Nobody voted on this. There was no democratic representation or investigation of the tradeoffs, only reflexive panic. Some people are like you and the thought of any of this never occurs. Some people are living within the fallout and see it clearly every day. The only problem is those latter people aren't media people and the media has gotten used to ignoring them for decades until they turn out to be convenient for an agenda. Just like every depressing little third-world disaster story full of ghastly (but vague and distant) statistics. Today it's just that the covid deaths are more useful than the ones being perpetrated by the people reporting the covid deaths. It's only natural that they don't want their blunder to come out; they are themselves culpable and trying to avoid accountability.

>>19037 >Won't work. You have to assume actual cases are at least four times larger than confirmed cases due to how some people exhibit minor symptoms. most of the people who get tested in my county show no symptoms at all. I use weekly cases because The sampling rate in my county is high enough that a high growth rate in real cases would be reflected by a growth in measured cases. My county is mostly rural so a pretty large chunk of people get tested. Improved models would try to estimate the exponential growth rate by looking at previous growth rates for a worst-case projection. I'm only interested in very short term prediction(as in, less than a week) and only interested in the worst-case so it doesn't need to be very complicated. >>19038 >tl;dr: don't fiddle and pretend you're qualified to make models. eh, I'm gonna, because I have the data and the scientific training and I might as well. Otherwise I'm a blind duck working based off my gut instinct and not fact, and we've already seen how that turns out when it comes to new diseases. It's also just an interesting technical problem that I like working on. >>19043 >And that doesn't even get into the philosophical question of whether we should allow an academic to have total power to override constitutions and enact global martial law—even if they turn out to be 100% correct. Nobody voted on this. Honestly, to me it seems like the panic was mostly from the people rather than the government or doctors, at least in the US. In the early days of lockdown, it enjoyed enthusiastic support from most of the US population. Even now "covid deniers" are treated as dangerous heretics by a lot of people, although I think the blind panic is starting to wear off. It may be different in Europe, it always is.

>>19044 >It's also just an interesting technical problem that I like working on. Fair enough, it's a free and open society after all. Just be aware your data training isn't going to prepare you for everything. I wouldn't trust simulation numbers much more than gut instinct. It's still not fact. Just something to keep in mind.

>>19044 >it seems like the panic was mostly from the people rather than the government or doctors Totally true up to a point; the panic always starts in the collective unconscious—so to speak. But there are constant moral panics bubbling and writhing beneath the surface of society. There's been plenty of others just this year alone, from BLM to Epstein and beyond. Ultimately it was government bureaucrats and pundits who decided which to seize upon and make actual policy. I don't remember people voting on closing down the whole service industry, for example. Bars and restaurants, local employers of tens of millions of precarious Americans. Maybe there were some vague legal mechanisms used to permit the decision constitutionally, but then again maybe not—power doesn't follow rules. It does what it can get away with when the populace is too cowed to hold it to those rules. It all just kind of happened. But there was an exercise of top-down power nonetheless. It wasn't all purely stochastic and driven by a populace who craves the boot, because a good portion of the populace always craves the boot. Anyway, getting off topic again so I'll cede the discussion back to models and data. Thank you for indulging all of this.

>>19046 Nobody cared while it was happening and things were silent for 3 full months after gov got the notification of a potential pandemic. People only started caring after media ran panic now pls whole day and even then media complained about people caring too little. People had to be coerced to care.

>>19048 Well. The media decided they had to be coerced to care. Packaged in that statement is the assumption that caring was warranted. Might have been a more complex call than that and they may not have had that right. Sometimes the homeostatic instinct in a crowd is revealed to be a useful feature, not a flaw. We are biased to perceive even unusual moments as a regression to the norm because the norm exists for a reason and is more stable. Sometimes it's a stubborn refusal to face reality and perhaps bystander effect, but skepticism is often a virtue.

>>19048 >>19049 Keep in mind we've ignored pandemics in the past. Even the ones we didn't ignore were never reacted to with this intensity. Not even close. Global lockdown and quarantine of the healthy is a totally new phenomenon. And there were still more that were totally undetected and went by stealthily until statisticians went over the death records the next year. Bodies rarely pile up in the streets. Not literally. I know it seems dangerous and heartless, but not reacting is still a strategic option. If emerging reports about non-covid excess deaths and future lockdown deaths are correct, it may have saved up to a hundred times more lives to just let it go. This is basically a great big trolley problem. We chose to take action and switch to a track with far more people tied to it, but there's uncertainty about how many are on each track so far.

I'll just leave this Bayesian regression here too. We have roughly a month more data by now, but afaik this has not changed much. Any conclusions that stand out? Also would be interested in knowing more about the specific methodology and source for this one. As far as I can tell it's real and a fairly credible study. Source mentioned in the graph, of course. I do realize there's an awful lot of variables to account for. But even so, more should be hoped for than this would seem to indicate.

>>19050 >it may have saved up to a hundred times more lives to just let it go. This is basically a great big trolley problem. We chose to take action and switch to a track with far more people tied to it Even if it is true that the disruption to the economy is more deadly, the real problem here is that we tied those people to that particular track to begin with. That track should be empty, and we have the option even now to empty it, by providing robust economic assistance to people regardless of their employment. The problem isn't the lockdown, it's the refusal to be flexible. This problem exists even without the pandemic, so it's probably overall a good thing that attention is being called to it in such a relatively mild way. We are going to get worse global disasters than this one in the future, and the sooner we realize that it isn't worth sacrificing peoples lives to maintain our consumerist lifestyles, the better.

>>19052 I see this rationale a lot, and I both get it yet I don't get it. We theoretically have the option to "empty the train," to be flexible about our economic system. But factions have been fighting since the first industrial revolution for that kind of economic justice. Centuries. We fought bloody battles for decades just to get 8-hour workdays down from 14 hours, weekends, and fire escapes on our child sweatshops. In the US. We still have that stuff everywhere else. I don't think we could slap together global UBI or whatever just yet. Though it's conceivable, it's not pragmatic. This type of hand-wave ignores political realities. We could someday build an economic system that doesn't slaughter us if we shut it down for six months. But we obviously couldn't have started and finished it all in the spring of this year. We all knew this. So we might have finally sounded the alarm for economic changes we desperately need—I'll certainly agree with that. But how callously we dismiss hundreds of millions as inevitable sacrifices. They weren't. We had problems with or without covid, of course that's true. But the scale of them now is so so much worse. We've seen precipitous drops not even matched by the great depression of the 1920s and 30s. And I'm talking real things like housing, family savings, employment... not the stock market. Market's doing great, apparently. Because the state is quick to restructure that economy and save their own. Every time. And we've seen the economy slaughter millions plenty of times in history. If we didn't learn then, forgive me for being so pessimistic about us learning now. If one is really fighting for the honest working people of this nation and the world, then we wouldn't be so careless with their lives. Even if it was for the greater good. Nobody acknowledges this greater good. It just feels like a way of plastering over our conscience for a while. I hope I'm wrong and they deliver aid to those who were prohibited by the state from feeding themselves, I really do. I've read my history books, though.

>>19053 And this still doesn't address the difficulties in building centrally-planned economies for a single nation, let alone the whole world. That was notoriously difficult and still led to things like the holodomor. We're seeing another holodomor crest over the horizon, in my opinion. It was just all so simplistic. I think we should be able to see by now that we're rationalizing post-hoc, not being a prudent and proactive culture.

>>19054 >That was notoriously difficult and still led to things like the holodomor. it really wasn't *that* much more difficult than capitalistic economics. Holodomor was political, not economic.

>>19045 All models are wrong, some are more useful than others. >>19051 interesting, whats the dead_per_M variable measuring? Just dead from coronavirus or total deviation from baseline? >>19050 This thread is for quantitative analysis. There are other threads for having these tired arguments.

>>19041 >>19043 I see. So would you happen to have any actual factual data about how much of the increase in deaths from all causes this year was a result of lockdowns? (Arm-waving and wall-of-text blathering don't count.)

>>19056 That axis refers to people determined dead from covid. Lemme see if I can find the original study again, I have it around here. Here's one that compares the progression of the disease across several locales, showing that once variables are controlled for the lockdowns don't seem to show any affect on the progression. Every country had wild variations in countermeasures and countermeasure timing, yet the disease tore through for about 30 days after the first 25 cases and went away just the same regardless of interventions: https://www.nber.org/papers/w27719.pdf

>>19060 now THAT is an interesting graph. Thanks lain, I'll check out this paper after work

>>19061 Would be interested to hear about your insights too

>>19064 really interesting article. I've noticed that the effective reproduction rate in my area has also dropped significantly even while people start relaxing their self-imposed isolation, which matches with the conclusion in this article. One possibility they didn't consider is that this could be due to improved treatment. But that shouldn't matter much since they considered some countries with very poor medical infrastructure like Mexico.

>>19065 What kind of improved treatment would you mean? Don't the vast majority of cases never require hospitalization or are completely asymptomatic? It would seem that most of the spread is driven by entirely untreated individuals who never get more than a test. But I could be fully wrong on that. Especially if asymptomatic spread is rare like the WHO said in early July and then did not ever actually retract, but "clarified"—at least partially under political pressure. But that would have changed the pandemic strategy in a far more profound way than anything else. Anyway, that's a different tangent. I've seen conjecture saying that it could be a combination of the harvesting effect (it killed the most vulnerable hosts almost immediately once they started putting poz patients back in nursing homes, a fantastic blunder) combined with some partial immunity incurred in the wild from different, older coronavirus strains... plus the revelation that herd immunity shows some signs of ramping up at a significantly lower population seroprevalence rate. Just speculation so far (though some parts like the lower herd immunity threshold have field data already) and of course there could be any number of other factors. The path a disease takes in a population has many variables, and top-down social intervention could only ever be a portion of the story. The question is whether it's a significant portion or just a little nudge among enormous, dominating natural forces. Obviously keeping people apart should slow it down according to common sense, but speculating about a single mechanism is so different from observing what really happens in the field with all known and unknown mechanisms showing up together. Especially natural human inability to follow unrealistic policies. I think it's craziness blame millions of random, unaffiliated civilians for the failure of overarching national plans. That's just craven; seems like a smoke bomb to let the incompetent bureaucrats slip away. https://science.sciencemag.org/content/sci/369/6505/846.full.pdf -One of the more current studies for herd immunity threshold. Another misconception I've seen in people is a strange idea that herd immunity would be binary; as in it would do nothing until that exact seroprevalence was hit, then it would be as if a vaccine had been teleported into everyone's bloodstream. I'm being a little hyperbolic, but obviously it must be analog: the effect ramps up—perhaps not quite linearly but to some degree—with a portion of that figure causing spread to gradually taper off the higher it gets. It would certainly have more of an effect than most eleventh-hour grocery store rules and such. So if correct, a far lower herd immunity finish line combined with higher than expected seroprevalence which some testing methods missed... plus a far more recent IFR and crude mortality figure is precipitously lower and seems to go further down by the day. It all dramatically changes the calculus for how many lives it would cost to quite literally ignore it away. I know that's perhaps unthinkable to the personal moral calculus of some, and I get that. But the heads of state should be singularly prepared to make hard decisions like this and recognize that every strategy is considered when we're examining the possible trade-offs. Because everything in the world has cost, and cost does mean lives.

>>19075 >What kind of improved treatment would you mean? Don't the vast majority of cases never require hospitalization or are completely asymptomatic? yes but in the paper they use deaths due to coronavirus, not number of cases, because they claim it is less noisy data(which I sort of believe). In this case treatment improvements would have a direct effect on the numbers.

>>19077 That's true, good point. Yeah I also believe it's vastly more reasonable to use deaths instead of "cases." Treatment probably does have some effect then; but still, the rationale for NPI like quarantine of the healthy and universal mask wearing is that it saves lives. This data would seem to show it doesn't; some other combination of variables must be at play, even if it does include some human actions. We're now seeing a lot of goalposts moved, like California's lockdown changing from keeping hospitals from being overwhelmed to keeping vague case counts down to keeping positive percentage down (a terrible metric, as any medical professional would tell you). Now they've finally arrived at the goal of "elimination," which is an impossible fantasy and is wildly different from the original rationale that these emergency powers were sold under. We've never eliminated one of these, and I think the image a lot of people have in their minds is this revisionist distortion of the struggle against smallpox or something of that nature. I would love to see data like this and critical counterpoints appear in the media. For some subjects, there is vast scientific consensus and thus it's deceptive to provide counterpoints just for the sake of "objectivity." I get that. But this is a rapidly-emerging situation where the data hasn't even completely stabilized; to act like there's broad scientific consensus already is just a dead giveaway that there is political pressure forcing that condition. Yet that's what people see in the news, and they think that is what "science" is. It's frustrating.

>>19075 >Treatment probably does have some effect then; but still, the rationale for NPI like quarantine of the healthy and universal mask wearing is that it saves lives. This data would seem to show it doesn't; some other combination of variables must be at play, even if it does include some human actions. Another issue is how often those guidelines actually get followed. In suburban America people might be afraid to go outside even without policy mandates, in Mexico City they probably figured it's not a big deal compared to starving and nobody cares what the government says anyway. I bet the correlation between low R_0 and wearing masks in public is a lot stronger than the correlation with never participating in any in-person activities, for instance. I saw some survey data about it recently but I read the questions and they were way too vague, not a good survey at all, I've been meaning to look for a new one and look at the correlation. >Yet that's what people see in the news, and they think that is what "science" is. It's frustrating. This is a longstanding problem. "science" has always been politicized, in the most extreme cases it has been used to justify genocide.

the state of CA just released guidelines calling coronavirus "widespread" in a county if there are more than 7 new cases per day per 100,000 that's ludicrous if you ask me, my county has sat at just above that number for months with no sign of exponential spread, in fact hospitalization numbers have gone down. This is also while people are starting to relax about it. I'm starting to wonder if there is some cut-off where exponential spread becomes dominant... would be interesting to investigate, but unfortunately it's damn near fuarrrking impossible to find good data in a machine-processible format publicly. Because of course it is, that data is valuable!

hello, I just thought I would explain a good way to draw conclusions from the crappy data often available online. A lot of the time, while decent raw data is not easy to come across, you can find data on the "doubling rate", which is the number of days until cases would double assuming current growth rates, usually smoothed. From this you can guess the exponential growth rate of case reports. For a variety of reasons this is much more reliable than number of infectious cases. The formula is r = 2^(1/N) where N is the doubling rate. Now for identifying exponential versus sub-exponential spread: log(r) is a measure of "how exponential" the spread currently is. Log(r) >= 1 is pretty much the worst possible scenario and it also corresponds to N < 1. Most of the time you will see log(r) between 0.1 and 0.001. A good way to see how log(r) corresponds to the degree of exponentiality is to look at which point the best-case and worst-case growth estimates (linear versus exponential) differ by a significant amount. Start with an estimate of the number of currently infectious cases C, and then find the t value when C*e^(log(r)*t) - C*(1+log(r)*t) >= 1, that is, when do we get a big enough difference that it matters? using the definition of the exponential function we can see that this will be the case when log(r)^2*t^2 / 2 ( 1 + log(r)*t/3 + ... ) >= 1/C. We can pretty much ignore the 1 + log(r)*t/3 + ... part, estimating it away to 2. So we solve for t^2 >= 1/(C*log(r)^2), so t >= 1/(sqrt(C)*log(r)). In my immediate region, with a C much higher than the official count, this is a little more than 19 days. So for now, growth is essentially linear. Why does linear versus exponential growth matter? Linear growth indicates that the disease is spreading very slowly and very predictably, and if growth is linear for a long time it means that you should not anticipate wild spikes. Overall I wouldn't say it is an indicator that there is no risk, but it is a good gauge of risk.