Coronavirus Sars-CoV-2/Covid-19 Megathread

[The Minsky Moment]

Florida is almost exactly middle of the pack on that measure.  Compared to Hawaii the state is a disaster.

[alfred russel]

Florida is almost exactly middle of the pack on that measure.  Compared to Hawaii the state is a disaster.

The US average is 375.6 deaths per million. Florida is at 151.6. If you want to remove the tri states area, the US total falls to 247, or still 62% higher than Florida.

[The Minsky Moment]

Florida is almost exactly middle of the pack on that measure.  Compared to Hawaii the state is a disaster.

The US average is 375.6 deaths per million. Florida is at 151.6. If you want to remove the tri states area, the US total falls to 247, or still 62% higher than Florida.

Your are using means not medians and these rates are censored left at zero (Hawaii is near zero).  Florida is middle of the pack - 27 states higher the rest lower.

[DGuller]

Minsky, to me it looks like you're all over the place, without having a point to make.  I think AR's point is simple, even if arguable.  "People say that Florida's response sucks, numbers say that Florida is not being hit as much as you would think if the response did really suck.  Maybe it didn't suck, or maybe our understanding of what's a sucky response and what isn't is flawed?"  I can easily see his point, unless I'm misinterpreting it. 

Now, you can argue whether it's more appropriate to look at where Florida is now (which is okay-ish), or whether you should be focusing more on the trend (which is not good), and that would be a meaningful discussion.  However, it appears to me that you're just reaching into a basket of statistical factoids, rifle through until you find somthing useful, and toss it out without having a thesis to support.  So what is your thesis?  Is it that Florida's response is a disaster, and that there is a statistical evidence to prove it?

[DGuller]

Your are using means not medians and these rates are censored left at zero (Hawaii is near zero).  Florida is middle of the pack - 27 states higher the rest lower.

BTW, I wrote the post above before seeing this.  I would not have been as complimentary of your statistical arguments if I saw this one.

[The Minsky Moment]

Right NOW Florida looks really bad. 
Eg New York with a smaller population is testing twice as many people per day as Florida but reporting 5 times as many new daily cases.

The reported death rates are still reasonable - at least if having a couple dozen people die every day is now considered reasonable - but there is an implicit assumption that the reported rates are properly capturing all COVID-19 deaths* and that those rates will continue to hold despite the sharp rise in cases.


*they don't for example, capture deaths in Florida by persons who maintain a primary residence out of state, a significant category in Florida.

[The Minsky Moment]

  Is it that Florida's response is a disaster, and that there is a statistical evidence to prove it?

My thesis is simple: we don't know yet.  But the jury is definitely out on Florida's "success"

[alfred russel]

Your are using means not medians and these rates are censored left at zero (Hawaii is near zero).  Florida is middle of the pack - 27 states higher the rest lower.

I don't see any reason to use means. There are a bunch of flyover states with low population densities, few people, and low covid 19 that will skew the analysis.

Florida is a highly populous and densely populated state--not far off in density from NY state. I'm willing to throw out the tri state area because NYC skews things, but comparing Florida to an island state like Hawaii or Wyoming is unproductive.

[DGuller]

By the way, "censored left" means that you're capping the rates at zero.  In other words, the rates could be negative, but you're rounding them up to zero.  From what we know about Covid-19 so far, I don't think it brings the dead people back to life. 

I guess what was meant was that the distribution of death rates is skewed, because it can't be negative.  Yes, it is, but so what?  AR's comparison of the average Florida death rate against nationwide death rate was completely statistically valid.  Claiming that Florida is average because it ranks near the middle on death rates among states is a statistical equivalent of a word salad.

[Admiral Yi]

My negative judgement of these currently spiking states comes from the fact that we have a great deal more information about transmission, fatality, etc. than NY did at the very beginning.

[The Minsky Moment]

By the way, "censored left" means that you're capping the rates at zero.  In other words, the rates could be negative, but you're rounding them up to zero.  From what we know about Covid-19 so far, I don't think it brings the dead people back to life. 

No matter how effective or efficient a state is at controlling coronavirus spread, they cant get below a 0 rate. In that respect it is similar to say purchases of durable household goods which can be zero but not lower, which happens to be the problem that motivated the original tobit model.

I guess what was meant was that the distribution of death rates is skewed, because it can't be negative.  Yes, it is, but so what?  AR's comparison of the average Florida death rate against nationwide death rate was completely statistically valid.  Claiming that Florida is average because it ranks near the middle on death rates among states is a statistical equivalent of a word salad.

If the point is compare one state to other states, then the logical way to do that is by a rank order of states (median).  Otherwise you have a calculation where the considerable majority of states are like the inhabitants of Lake Wobegon - all safely well above average.

[DGuller]

By the way, "censored left" means that you're capping the rates at zero.  In other words, the rates could be negative, but you're rounding them up to zero.  From what we know about Covid-19 so far, I don't think it brings the dead people back to life. 

No matter how effective or efficient a state is at controlling coronavirus spread, they cant get below a 0 rate. In that respect it is similar to say purchases of durable household goods which can be zero but not lower, which happens to be the problem that motivated the original tobit model.

None of this is relevant.  Most useful distributions in real life have a domain that starts at zero.  This is has nothing to do with them being left-censored at zero.  It's a minor point, I only pointed it out because it seemed like you were throwing some technical term randomly to make it look like it supports a dubious assertion.

I guess what was meant was that the distribution of death rates is skewed, because it can't be negative.  Yes, it is, but so what?  AR's comparison of the average Florida death rate against nationwide death rate was completely statistically valid.  Claiming that Florida is average because it ranks near the middle on death rates among states is a statistical equivalent of a word salad.

If the point is compare one state to other states, then the logical way to do that is by a rank order of states (median).  Otherwise you have a calculation where the considerable majority of states are like the inhabitants of Lake Wobegon - all safely well above average.

Which is perfectly fine.  In any given year, 90%-95% of drivers are above average when it comes the number of crashes they have.  That's how real life works sometimes.  A median driver has zero crashes, which is probably not the most helpful number to understand how often people crash. 

I think most people care about their odds of dying, rather than how their odds rank among states.  By your logic, if 99.99% of Covid deaths happened in NY, then you'd still have 24 other states above average in risk, even though the people's chance of dying there would be essentially zero.

[viper37]

I like that the new distancing rule in England is "1 meter plus". The entire government is in that line how they couldn't just say "1 meter".

I do not understand why we're having such an arcane, scholastic debate about this that's somehow gone on for weeks.

Also the approach of the different nations who are all saying they're sticking with 2m. My prediction is after making a great deal of hay about how risky the English government has been on this, they'll also adopt 1m or 1.5m or whatever in about a month's time. Just like the Scottish government saying no-one understands the English slogan of "stay alert" because it's so vague before, a month later, announcing their slogan: "stay safe"

I see all this political point-scoring over what seem like quite small points and feel re-assured. Nature is healing.

the director of Quebec' Public Health was asked that question a while ago:
"Why 2m here when in many European countries they are at 1m distancing?"A: "Because when I say 2m, people are at 1,5m.  If I say 1m, they'll ignore the rule".
But we've already reduced the distance in many places to 1,5m, 1m (teens) or 0m for very young children.

[The Minsky Moment]

None of this is relevant.  Most useful distributions in real life have a domain that starts at zero.  This is has nothing to do with them being left-censored at zero. 

Plenty of relevant and useful distributions aren't cutoff at zero: wealth can be negative (debt), net income can be negative, economic growth can be genitive, inflation can be negative, among many others. Some are cutoff at zero and its pretty common to take that in account when performing an analysis on the data set, such as through the use of a tobit. It can be misleading to using a straight mean when dealing with a variable that is bounded at zero but unbounded in the other direction. 

It's a minor point, I only pointed it out because it seemed like you were throwing some technical term randomly to make it look like it supports a dubious assertion.

I'm glad we resolved that mistaken concern

Most people care about their odds of dying, rather than how their odds rank among states. 

I can't say what most people care about, I haven't surveyed them.  But I wasn't responding to most people, I was responding to AR who was making a claim about Florida having a good outcome *compared to other states*   In addressing that claim, the rank among states is not only relevant, it is really the only relevant criterion.

[viper37]

Right NOW Florida looks really bad. 
Eg New York with a smaller population is testing twice as many people per day as Florida but reporting 5 times as many new daily cases.

The reported death rates are still reasonable - at least if having a couple dozen people die every day is now considered reasonable - but there is an implicit assumption that the reported rates are properly capturing all COVID-19 deaths* and that those rates will continue to hold despite the sharp rise in cases.


*they don't for example, capture deaths in Florida by persons who maintain a primary residence out of state, a significant category in Florida.

do they captude deaths of people not in a hospital, like in a private or public elderly's residence?  It seems some States only considers deaths at hospitals, and/or only from patients positively tested to covid-19.