green03

Members
  • Content Count

    291
  • Joined

  • Last visited

Community Reputation

17 Whiner

About green03

  • Rank
    Anarchist

Recent Profile Visitors

The recent visitors block is disabled and is not being shown to other users.

  1. green03

    History of Red-on-Right-Returning

    Sure, the theory is all there, but if you never encounter one day to day, it's hard to convince the brain to hold onto that "useless" knowledge. They clearly seem to be deployed differently in different regions. Anyone know where they would be most commonly found? (Highest density or highest ratio among marks)?
  2. Reading past the long, fruitless back and forth. We can be confident of one thing: the numbers reported for China are a best possible scenario. With the obvious interest to look good, it's a near certainty that any deviation from actual is never in the direction of more reported cases/death than what actually occurred. We can also assume that few places have the will, social cohesion and control to replicate the stringent measures taken by China. This is particularly doubtful about places that strongly value individual gain over collective responsibilities. Third item worth noting (again) is that of all the published measures anywhere, deaths are a relatively reliable metric (although some countries are rumored to be exceptions even there). High mortality rates may therefore reflect a high rate of missing cases, and not so much regional differences in susceptibility - although there may be a bit of that. Fourth: the growth rate (as opposed to the actual daily count) is less susceptible to systematic errors in measurement, because those errors would have to change exponentially over time. If you constantly miss 50% of your cases, your totals are off by 50%, but because of the nature of exponential growth, your growth rate is the same as for the actual (though unknown) number of cases. The good new for that is that you will be able to tell a "flattening of the curve" even with an imperfect testing protocol -- as long as your tests have a good portion of negatives, that is. Otherwise your number of tests performed would directly limit the number of cases detected, and the curve collapses to reporting the testing effort, not sampling the infections. Putting that all together, you can go to a site like: http://www.91-divoc.com/pages/covid-visualization/ and "splice" a "Chinese curve" or "Korean curve" onto whatever country's data to get a good estimate of what a likely "best case" scenario would be going forward. For the splicing, you'd have to match parts of the curve with similar growth rate, not similar total cases, or "days since infection". Flattening the curve starts with your existing growth rate and goes from there. The official (best case) curve for China, was rising 1.22 on day 10. As is the US curve on the last day plotted here. If the US were (warning: counterfactual hypothesis) already engaged in the type of actions China was engaged in 1o days after their 10th death, then a best case scenario might be if the US curve for future days followed the Chinese curve from day 10. The Chinese curve continues to grow while flattening, from 259 to 3300 deaths or a factor between 12 and 13. However, the US deaths stand at 2467 in the same version of this data set. Therefore, a "best scenario" for the US is by now upward of 31K. Unfortunately that is so optimistic as to appear completely out of reach. Both because the Chinese data themselves are likely optimistic, and because there's no indication the US is anywhere near to having instituted effective measures that are widespread enough. (Although some states already look better than others). Feel free to play other scenarios, like what if the US can move onto an ROK-like curve? That curve is still far from really flat, and Korea (like Japan) manage to get onto it rather from the outset - making it less likely to be a model. The Italian example may not be as comparable, because the death numbers are so much higher per population - so you might run into saturation effects. Saturation effects would start to apply even before you get herd immunity and are a reason why even the worst case scenarios stop short at something like 50% of the total population not infected: at some point, there are so many active infection sources that multiple sources are infecting the same target, which is different from herd immunity where sources can only infect a portion of the exposed, because the remainder, while present, is no longer susceptible.
  3. That. An "invisible halo" of mild cases can't grow faster than the severe "visible" cases; by necessity it follows the same exponential, and therefore if a certain fraction of cases is too mild to detect with current testing protocols or sampling methods that fraction would remain the same. Mortality rates limit how long the virus can be undetected in the community: those deaths will show up. Realistic estimates for the "halo" are much lower than what ppl with wishful thinking are brandishing about. And the range of uncertainty for the exact case mortality rates don't provide enough cover for the level of speculation. The numbers don't work out. However, I have no illusion that wishful thinking can be influenced by facts.
  4. http://www.91-divoc.com/pages/covid-visualization/ Look at the data normalized for population. Several European countries are today at about 1000 cases/1M. The US is less than a week away from that number (and not likely to hit the breaks fast enough to slow things down before then). In fact, it's likely that growth will continue for another week, to 10 000 cases/1M. At a 2% case mortality you have 200 deaths / M or 60 000 - that's the bottom rung of the estimate, and assumes very strong measures to slow / halt the growth. With no prior immunity the peak in total cases would be more like 50% or 500 000 / 1 M. or 10 000 death / 1 M for a total of 3M death. A far cry from a typical "flu". With immediate, focused, and effective actions to contain/slow this thing, we would have been able to keep this below the level of a "mild flu". At this point, the best case scenario will look like a bad flu year - in a month and a half (not spread out over a full season) - continued inaction will get us to that worst case scenario. All the states for which the numbers look like they are headed in the right direction have been ramping up countermeasures for the last few weeks. Look at the WA State numbers in the source I gave, to see that actions taken from March 8, are now paying off. Slowly, not fast enough, but somewhat.
  5. Unless the testing protocols are the same you cannot compare absolute numbers across countries. I think we all agree on that. But I believe you can compare growth rates. At least I've not come up with any reasonable scenario how testing can make the data look like a different growth rate. Not over any significant time span. There are some extreme scenarios like testing so few people that all tests are positive. At that point, your curve maps your testing activity, and is independent of the actual number of cases. It's more common that your tests will miss some fraction, but that fraction would have to grow exponentially to mimic a lower growth rate. That's not the most probably scenario. It's more likely you test a similar risk pool and perhaps mostly confirm cases of a certain severity. But that should be a rather constant fraction of the total case load, and by the nature of the exponential growth, the rate for the fraction is the same as the rate for the total. And that should work both within the same country (comparing different time periods) and across. Look at the visualization and let me know if you think that normalizing to population changes the trends in any significant way. There are both sets of curves. It doesn't, because the population is a constant denominator, therefore not affecting the growth rate. http://www.91-divoc.com/pages/covid-visualization/ Absolute numbers tell you whether you have 1M or 10M cases in the end (or at any given day). The growth rate shows you how fast you are going to get there. 1.35 daily growth is a factor 10 about every week or so (8 days). In a month and a day, that's a factor 10 000. Let's hope more countries/territories/states start flattening those growth rates. To double the period from 8 to 16 days, requires the rate to be 1.15, or 1.07 to get 32 days.
  6. http://www.91-divoc.com/pages/covid-visualization/ One of the best visualizations I've seen. Shows log plots to be able to see change in growth rate, and linear, for those that just want to see the steep curves. Also has figures normalized to populations and gives both cases and deaths. For the US, look at the WA State numbers. There's a distinct flattening of the straight line in the log plot, meaning that for about a week the growth rate has been lower. (You can make some arguments about testing, but it appears test coverage would have to be getting exponentially worse to simulate a lower growth rate; constantly missing a similar fraction of your cases would shift, but not bend your curves in the log plot. Testing inadequacies can affect our knowledge of the overall cases differently from knowing the rate of growth, because if you only discover a constant say 10% of the cases, the growth rate, and the time to double, is unaffected - that's just the nature of exponential growth). So, there are indications that this flattening is real. Incidentally, it shows up a week after the bigger employers in the epicenter for the state told people to work remotely. (And the first restrictions on larger public meetings). Now, the state has stepped up the lockdown, will know in a week whether the numbers respond. If millions had quietly acquired immunity out of sight, you'd see the exponential turn into a logistics curve (see Wikipedia). That's the S curve that first rises sharply and then flattens out rapidly as the growth is limited by resources (non-immune people). Looking at the data set linked above, I don't see any evidence of that; most data points for most countries are still well fitted by 35% growth exponential.
  7. green03

    Covid-19

    http://www.91-divoc.com/pages/covid-visualization/
  8. What is the current growth curve at 35% increase daily? Something like a factor of 10 every 8 days. In just about two weeks you can expect 400 to turn into 40,000. And a little over a week after that, you find yourself at 400,000, or anywhere from 7 to 35 times above flu level. I don't think of 3-4 weeks as such a long time. http://www.91-divoc.com/pages/covid-visualization/
  9. A lot of good stuff has been written on the fact the denominator isn't known for the mortality rate, because only confirmed cases can be factored in. But, you know what, in exponential growth it does not matter. The daily percentage increase in deaths is not dependent on the mortality rate being known, because it would be the same whether these deaths were due to high mortality for a small base of infections, or lower mortality rate, for a larger base. Currently the daily increase is about 33%. That seems to be the same for all countries with uncontained/unrestrained exposure. That means, the absolute number of deaths will grow at that rate, which comes to a factor of 10 every so many days (less than two weeks). After another equally short period, you have a total factor of 100. -- as long as the social distancing and whatever do not result in a lasting change in that daily increase (or until the virus runs out of population altogether). The same goes for the number of severe cases that do need care but don't result in death. We don't need to know the rate per infection to know that the absolute numbers will grow exponentially and if they are today twice as many as the deaths, they will continue to be twice as many when there are more deaths. So, really, to see when the health care system will be overwhelmed, you only need to know how many cases of a given severity there were yesterday, and how many there are today. Then run the growth curve forward. If testing cannot be deployed at a scale where it can change people's behavior in time (isolation/quarantine) knowing the real number of mild(er) cases won't help with the hard problem. The only "benefit" of a large halo of milder cases is that the virus will run out of infectable population a bit sooner -- the likely ranges for those estimates won't make a difference if we continue to have a growth curve with 33% daily increase -- given the likely range of these estimates, that total number will have more severe cases than we can handle / could handle, even with a perfect health care system. That's about all we need to know about knowing the denominator: it's looking rather academic.
  10. Looks like the chance to contain this, if there ever was one, is probably gone. Uncontained, the cases will rise exponentially, until the virus runs out of uninfected people. The game now is to "flatten the curve": if you can slow the spread, you have fewer people that are sick simultaneously, the better the healthcare system can cope with the critical cases. (From the data in post #28 it looks like Japan may have been successful in that). The "normal" flu, spread out over an entire season may be "baked in" to our expectation and what the medical system can cope with, but this one looks ~20 times more deadly, and may peak much more quickly. That, and not the small current case numbers, is what all planning and counter measures need to be directed at.
  11. green03

    I'd like to thank the previous owner for...

    For fixing a hole below the waterline with caulk.
  12. I enjoy a good political discussion, and no, that does not mean everybody nodding in agreement. Differences in political views, and more specifically, differences in views on specific policies are one thing. Being hateful is another matter. "If they don't have the money to buy health care, we should let them die", was apparently not an over the top joke in bad taste, but a sincerely held belief. I've not been on that boat since.
  13. That depends on the jurisdiction. It's definitely an issue in some locations. And, where applicable, your choice is to go dry or to find a way to formally sign up frequent guest as members. You still need to find some way to couple that with some benefits beyond simply being crew. If it's a license issue, and the venue is attractive on other occasions than race days, a natural benefit would be to allow unrestricted visits to the bar on any day (and ability to sign in an occasional bar guest). Conversely, you'd make crew who don't want to use the bar complimentary. If it's a licensing issue, people might be more understanding than if it's merely money grabbing. If practical, sign people up for that at the bar, not at race registration.
  14. I'm member in a club that instituted "bar memberships". Not a sailing club, and the main motivator was to change the ratio of members to guests for purposes of the bar license. But in that case, there's a clear benefit for these second-class members: they get to use certain facilities of the club w/o being signed in by a full member. If you bring the same guest (and I believe all classes of members may bring guests) more than once, there's a strong push to have them sign up as member (of either class). For the club, they become a source of revenue (mainly at the bar) and a target for conversion to full membership. Not sure how much that translates to "crew memberships" where the only / main benefit may be to be allowed to work on someone's boat so that they have a chance at the pickle dish??
  15. green03

    we call bullshit

    Why is there a stage blocking the view of the water?