и немного статистики

[ny_quant]

Не умею вставлять из ФБ так что кликайте если хотите посмотреть на сравнительную таблицу смертности по ср. с гриппом 2018 года.

https://www.facebook.com/photo.php?fbid=10157077265145318&set=a.10151959543410318&type=3&eid=ARAVdnqjL68etGhoS9_KGMz9tlKSGCLuBIuT36KfKh0kaiPToWTcMCsxMk5cofAR5a8AGQ8vHV41yP9j&__xts__%5B0%5D=68.ARCKwiTJIxzG7Zf-xx6y0nNzFoYq4Hz3daImKxmfxsABsdm8VsYO7WRXxgmYenJ7F6n9789AbcVjz4u8w_9cQDBSRrTeu-s4Md2G3AlVBqzNLhe_I0GccsvRVyPaf5CcQk8VJrCHIM_Sy2yeZ3QQDB6tXhOYwjO2DzZndEKhNbDaYmS_lteNWsVjFlckbINnhIohIQcLOqtd9T76aQRRdlj8XfJ1wVsi186Vx-VA-lEU7fVBQALgwSdzD7XA31bJ4XFBWaJHuWqB_JEaCd9_9f8PaPh8OA8OWPZioJ8OdFaOdZqqpc4&__tn__=EEHH-R

Comments

[misha-b.livejournal.com]

The flu is roughly at 30k (and that was a bad season, if I remember correctly). Coronavirus is already twice as bad after about a month. Even if it flattens now, it is going to be 5-10 times worse at least.

[ny-quant.livejournal.com]

Посмотрим. I'm on record predicting 100,000 deaths give or take a factor of 2 before anything started in US. Официальный прогноз сегодня: 74,073
https://covid19.healthdata.org/united-states-of-america
что меньше чем смертность от гриппа в 2018. Это конечно нечестное сравнение яблок с апельсинами, но всё же.

[shvarz.livejournal.com]

Как вообще можно этим предсказаниям верить если у них прогнозируемая кривая смертей выглядит симметрично до и после пика? Хоть в одной стране такое наблюдается? Ну, может быть более-менее в Китае, хотя и там не очень. После пика смерти на спад идут очень медленно.

[ny-quant.livejournal.com]

Я вообще-то не говорил, что я им верю. Если вы можете подсказать какaя модель сегодня считается consensus best, то я буду смотреть на нее.

[shvarz.livejournal.com]

Не знаю какая лучшая, но точно не эта.

[ny-quant.livejournal.com]

Может они все такие?

[shvarz.livejournal.com]

Скорее всего. Моделировать можно естественный процесс который протекает сам по себе, а тут столько всего зависит от принимающих глобальные решения и от поведения людей. Грубо говоря, завтра весь карантин отменят и новые инфекции скаканут вверх, какая модель может такое предсказать?

[ny-quant.livejournal.com]

Это совершенно справедливо.

[misha-b.livejournal.com]

I think your numbers are a bit off -- flu deaths in 2017/2018 season were about 61000 (https://www.cdc.gov/flu/about/burden-averted/2017-2018.htm#table1).

74000 for the whole year sounds absurd. We are at slightly under 2000/day right now, so we will be there in less than 10 days at the current rate.

[ny-quant.livejournal.com]

These are not *my* numbers and this is not my forecast either.

[misha-b.livejournal.com]

> что меньше чем смертность от гриппа в 2018.

You said that, no?

[ny-quant.livejournal.com]

Раньше они говорили, что 80,000:

https://www.statnews.com/2018/09/26/cdc-us-flu-deaths-winter/

[misha-b.livejournal.com]

I see. I guess they revised the numbers down.

[kodak2004.livejournal.com]

95% UI 46,404-94,987

80 тысяч укладываются в него и сейчас.

[alexanderr.livejournal.com]

so, it's just like flu, right?

[ny-quant.livejournal.com]

No, not exactly.

[alexanderr.livejournal.com]

what is the key difference? coronavirus death rate of 0.1% is similar to flu, no?

[ny-quant.livejournal.com]

The key difference is very strong dependence of mortality on age.

[alexanderr.livejournal.com]

is that why there's never a flu quarantine? different mortality by age, while similar total mortality?

[ny-quant.livejournal.com]

No, that's because regular flu ends on schedule and this one seems to be not.

[alexanderr.livejournal.com]

covid-19 data on nyc.gov shows very clearly that this is over.
at least in NYC. cases, hospitalizations, deaths, everything is back
to mid-March levels, i.e. low

[ny-quant.livejournal.com]

To the extent that it is "over", it is so largely due to the measures taken. As opposed to the regular flu that goes away on its own.

[alexanderr.livejournal.com]

officially 25% of NYC residents were exposed. that's based on random testing, looking for antibodies.
so with measures we get 25%. without measures we get 100%. was it worth it?
more importantly, do we need to continue with the measures? what would be the point of that?
unofficially, I think the real number is closer to 50%. since they tested only active adults,
no children, but included children in the total population. plus, the tests are not that accurate yet.

[ny-quant.livejournal.com]

Ответы на вопросы зависят от выбора метрики оценки результатов. Каковы допустимые экономические потери за одну спасенную жизнь?

[alexanderr.livejournal.com]

we could ask the same question differently: ~500 people losing jobs for ~1 person who dies, that is what we have currently. how many were saved and for how long, we'll never know. but record unemployment will destroy many lives, we know that very well. a great number of people who lost jobs will die, that one is easy to predict. based on historical data, if US unemployment hits 32% that would increase the death toll by 77,000. just as a direct result of losing the job. same order of magnitude as total virus deaths. so, it is not really clear if lives are going to be saved or lost