CoronaBro Meltdown/SARS-Covid-19 Spitballing Thread

[sys]

you're growing the denominator every day while the numerator is constrained by a semi-unrelated number (tests).  the number is obviously going to fall every day unless testing increases at the same rate or faster than does the incidence of the virus.

[sys]

https://twitter.com/michael_nielsen/status/1246457825720553472

[Justwin]

Is that day over day increase? Because 12% of 113,000 is more new cases than 13% of 100,000....

yeah, i don't know why people use those metrics.  obviously misleading/difficult to interpret.

Seems a decent way to show progress imo. I guess you could have pause if regions aren't taking similar precautions and what you're seeing is NY is driving the flattening while some other states are on a different path.

I also agree that cases probably have more noise than deaths.

I agree that cases can have more noise than deaths.  I do think that cases is a better metric to look at though for a number of reasons:

1.  Cases tell us where we're headed and deaths tell us where we've been.  This is true even (especially) as mortality rates may increase due to an overwhelmed health care system.  If you were going to guess where we'll be in 6 weeks, you're better off basing your estimate on cases and then extrapolating deaths off of that.  The trend in deaths is similar to that in cases, albeit with a lag.

2.  Our capacity for detecting cases is expanding every day.  Using the COVID Tracking Project's number, we had 216,000 test results in the latest 24 hour period (4:00 p.m. on 4/3 to 4:00 p.m. on 4/4).  That is basically the same number of results we got total from March 4 - March 22 and it is the highest number of any day.  Our capacity for detecting deaths is not really changing at all.  If anything then, the cases number is relatively pessimistic compared to the deaths number.

3.  Overall, the vast majority of the population is under some type of stay-at-home order.  I'm pretty sure it's over 90% of the population.  So while New York is probably driving the current slowing (due to their share of cases), I don't think the rest of the nation is going to explode and overwhelm that.  I would venture that most states, counties or cities implemented stay-at-home orders before New York did on a relative basis.  For example, Kansas shut down schools on March 17 while New York did on March 18.  Once it's all said and done, New York (as a state) will probably account for around 1/6 of all deaths.

4.  The daily deaths percentages are not following a different pattern than the daily cases percentages.

[michigancat]

https://twitter.com/michael_nielsen/status/1246457825720553472

Good contrast to Musk.

https://twitter.com/kenklippenstein/status/1246265058839191554

[Justwin]

you're growing the denominator every day while the numerator is constrained by a semi-unrelated number (tests).  the number is obviously going to fall every day unless testing increases at the same rate or faster than does the incidence of the virus.

The denominator is related to previous testing capacity.  If you're growing testing capacity, you're only growing the numerator without the ability to grow the denominator.

[DaBigTrain]

I just did a 100% tip on a DoorDash order. I still don’t feel like I’ve done enough. 

[DaBigTrain]

Also, should I have said this? Will I be put down for not doing more?

[sys]

Using the COVID Tracking Project's number, we had 216,000 test results in the latest 24 hour period (4:00 p.m. on 4/3 to 4:00 p.m. on 4/4). 

california dumped like 70k negative test results today.  almost certainly mostly older tests that just hadn't been reported, not tests completed today.

[Justwin]

Using the COVID Tracking Project's number, we had 216,000 test results in the latest 24 hour period (4:00 p.m. on 4/3 to 4:00 p.m. on 4/4). 

california dumped like 70k negative test results today.  almost certainly mostly older tests that just hadn't been reported, not tests completed today.

Even if you just throw out those 70,000 negative results, we still had more test results today than any other day.

[sys]

yeah, 140kish, similar to yesterday.  after being stalled out at 100-110k for a week or so.

[Justwin]

yeah, 140kish, similar to yesterday.  after being stalled out at 100-110k for a week or so.

If you're just going to throw out delayed results that were included today, you need to do the same for other days that had delayed results included in them.  If you do that, those days' tests results are going to decrease and the numbers for today will be higher even more on a relative basis.  I get criticizing data for anomalies in the way it's reported, but when you make corrections for one observation, you have to go back and do the same for all of the other observations that are subject to the same errors.

In this case, California has been slow to report a lot of results and has had a lot of pending results.  Their pending results only went down 46,500 today.  How are you correcting for the fact that there were 70,000 negative results dumped, but only a 46,500 drop in pending results?  If the answer is California is not a reliable data provider, what do you do then?

Your argument is basically that testing is increasing, but not increasing as fast as the data say it is?

[sys]

you could go back and distribute those 75k (it was 75k, not 70k as i previously said), but you'd just be guessing as to when they should go and i don't really see the point.  it is not ambiguous, however, that ca dumped a bunch of data today and it an event that is unlikely to be a reoccuring feature.

https://twitter.com/COVID19Tracking/status/1246559828073308160

[sys]

pretty good 538 article on testing regimes and how they impact reported cases.

https://fivethirtyeight.com/features/coronavirus-case-counts-are-meaningless/

[Justwin]

I am interested in when we are going to peak and start going down.  The actual number of cases is not that important to me.  That is why I like the day-over-day percent increase.  It tells you what is happening to the slope of the number of detected cases on a logarithmic scale.

In the 538 piece you linked, in no model was there a situation where the slope of actual infections was steeper than the slope of detected cases (perhaps very slightly in the testing capacity doesn't scale up model).  All that happened was that detected cases understates the number of actual infections and the peak of detected cases lags behind the peak of actual infections.

If the rate of increase is slowing for detected cases, then the rate of increase for actual infections is slowing and probably started slowing a week or two ago.  Whether California dumped 75,000 cases really has no bearing on this.  As I said before testing capacity is increasing (even if it is slowly and "stalled" as you say).  A slowdown in the rate of increase of detected cases is good news and a sign we're headed in the right direction.

[steve dave]

Yes, the stay at home orders should pay off big time hopefully


Sent from my iPhone using Tapatalk

[catastrophe]

https://twitter.com/michael_nielsen/status/1246457825720553472

Holy crap this guy. I need to start donating to the Gates foundation.

[catastrophe]

I just did a 100% tip on a DoorDash order. I still don’t feel like I’ve done enough.

You’re not gonna prop up this economy yourself regardless. Just eat out a couple times a week and tip reasonably and you’ll be doing your part.

[Justwin]

I am interested in when we are going to peak and start going down.  The actual number of cases is not that important to me.  That is why I like the day-over-day percent increase.  It tells you what is happening to the slope of the number of detected cases on a logarithmic scale.

In the 538 piece you linked, in no model was there a situation where the slope of actual infections was steeper than the slope of detected cases (perhaps very slightly in the testing capacity doesn't scale up model).  All that happened was that detected cases understates the number of actual infections and the peak of detected cases lags behind the peak of actual infections.

If the rate of increase is slowing for detected cases, then the rate of increase for actual infections is slowing and probably started slowing a week or two ago.  Whether California dumped 75,000 cases really has no bearing on this.  As I said before testing capacity is increasing (even if it is slowly and "stalled" as you say).  A slowdown in the rate of increase of detected cases is good news and a sign we're headed in the right direction.

First statement that I bolded...really...the history of this thread says differently and in general I don't understand how it isn't important if you care about actual outcomes and affected people. I guess you can say you said cases and not deaths (which to me is a value judgement as I don't know long terms effects), but they're correlated to the point where more of one is more of the other.

I agree that it's a nice stop light indicator, but it's only telling a tiny portion of the situation, imo it's are the social distancing measures effective.

We're far past deciding whether we're going to have stay at home orders or not, so right now it's not really worth it to me to debate whether we should have them or not.  There are some states that haven't implemented them, but even in these states, there are localities that have implemented them and I'd say easily over 90% of the population is covered by them.  Once this is all over, I think it will be interesting to go back and think about what the actual number of cases were, what the mortality rate was, what was the economic cost of the shutdowns, what would have been the economic costs without the shutdowns, etc. and think about the decisions that were made and when they were made.  So, yes, I am interested in the actual number of cases, but not really at this moment.

Given that we're shutting things down and we're not really debating it, I'm now more interested in looking at when things are going to start improving and what is a reasonable time frame for lifting the shutdowns.  Actual numbers might be important when we're deciding to open things back up, but it's likely that when we are willing to open things up, the number of detected cases will outnumber the number of actual cases due to false positives.

[Bqqkie Pimp]

https://twitter.com/michael_nielsen/status/1246457825720553472

Enjoy your microchip.

#ID2020

 

[michigancat]

I just did a 100% tip on a DoorDash order. I still don’t feel like I’ve done enough.

You’re not gonna prop up this economy yourself regardless. Just eat out a couple times a week and tip reasonably and you’ll be doing your part.

One way to help us to call the restaurant and see if they'll deliver themselves. Lots of restaurants have had servers become delivery people. It's a pain in the ass to order over the phone but it gives the restaurants a bigger cut than doordash, and a few aren't even on door dash.

[Bqqkie Pimp]

https://twitter.com/MichaelCoudrey/status/1236736335794167808?s=20


https://twitter.com/MichaelCoudrey/status/1246683885401804801?s=20

[treysolid]

I am interested in when we are going to peak and start going down.  The actual number of cases is not that important to me.  That is why I like the day-over-day percent increase.  It tells you what is happening to the slope of the number of detected cases on a logarithmic scale.

In the 538 piece you linked, in no model was there a situation where the slope of actual infections was steeper than the slope of detected cases (perhaps very slightly in the testing capacity doesn't scale up model).  All that happened was that detected cases understates the number of actual infections and the peak of detected cases lags behind the peak of actual infections.

If the rate of increase is slowing for detected cases, then the rate of increase for actual infections is slowing and probably started slowing a week or two ago.  Whether California dumped 75,000 cases really has no bearing on this.  As I said before testing capacity is increasing (even if it is slowly and "stalled" as you say).  A slowdown in the rate of increase of detected cases is good news and a sign we're headed in the right direction.

First statement that I bolded...really...the history of this thread says differently and in general I don't understand how it isn't important if you care about actual outcomes and affected people. I guess you can say you said cases and not deaths (which to me is a value judgement as I don't know long terms effects), but they're correlated to the point where more of one is more of the other.

I agree that it's a nice stop light indicator, but it's only telling a tiny portion of the situation, imo it's are the social distancing measures effective.

We're far past deciding whether we're going to have stay at home orders or not, so right now it's not really worth it to me to debate whether we should have them or not.  There are some states that haven't implemented them, but even in these states, there are localities that have implemented them and I'd say easily over 90% of the population is covered by them.  Once this is all over, I think it will be interesting to go back and think about what the actual number of cases were, what the mortality rate was, what was the economic cost of the shutdowns, what would have been the economic costs without the shutdowns, etc. and think about the decisions that were made and when they were made.  So, yes, I am interested in the actual number of cases, but not really at this moment.

Given that we're shutting things down and we're not really debating it, I'm now more interested in looking at when things are going to start improving and what is a reasonable time frame for lifting the shutdowns.  Actual numbers might be important when we're deciding to open things back up, but it's likely that when we are willing to open things up, the number of detected cases will outnumber the number of actual cases due to false positives.

Here's why % increase is a bad metric: to the uninformed person, it can easily appear as if the # of new cases is decreasing, when the opposite is actually true. This can affect people's behavior. If they feel like the tide has turned, they might be more lax about social distancing and other preventative measures like washing their hands, etc.

I'm a scientist. I deal with data all the time and see the ways that it's used to trick people. If what you care about is the the rate of new infections, then use the slope and report the doubling time instead of % increase.

[catastrophe]

I just did a 100% tip on a DoorDash order. I still don’t feel like I’ve done enough.

You’re not gonna prop up this economy yourself regardless. Just eat out a couple times a week and tip reasonably and you’ll be doing your part.

One way to help us to call the restaurant and see if they'll deliver themselves. Lots of restaurants have had servers become delivery people. It's a pain in the ass to order over the phone but it gives the restaurants a bigger cut than doordash, and a few aren't even on door dash.

That’s a good point. There are a number of restaurants that have converted their servers to delivery drivers. A lot do still use DoorDash though.

[cfbandyman]

Italy really seems to have turned the corner, seriously.

[Bqqkie Pimp]

And so is the CDC for that matter. 

https://www.westernjournal.com/cdc-tells-hospitals-list-covid-cause-death-even-just-assuming-contributed/


Fauci is a Bill Gates puppet and more interested in selling everybody a rough ridin' vaccine while the rest of the world is finding treatments and cures that actually work.   

Prove me wrong.