Good news on the case growth trend.

[Ms. Toad]

An 8-day decline in the rate of increase in new cases (with one outlier, which could easily be a report of cases that should be on 3/26 that were made before the close of worldometers day.

Daily multipliers (# cases on prior day x multiplier = #new cases for the current date)from 3/22 through 3/29:

3/22 - 1.385797497
3/23 - 1.303732189
3/24 - 1.253526924
3/25 - 1.244204075
3/26 - 1.252510592
3/27 - 1.218774507
3/28 - 1.186812131
3/29 - 1.149452168

No guarantees it will continue - but this is a nice long streak of slowing the growth. We're still seeing more cases each day (the multiplier is more than 1). But were not seeing as many new cases, proportional to the existing cases, as we have been.

I would expect a correction, at some point, similar to the one in China on 2/12 & 13 once we start to see the rapid tests deployed. But I am hopeful that this represents a real flattening of the curve.

[LisaL]

1. Considering how hard it is to get tested, I wouldn't count my chickens just yet.

Have people here actually tried to get tested? If a lot of people can't get tested, what good is the supposed slowing?

[Ms. Toad]

2. You did read my last sentence, yes? n/t

[LisaL]

3. I am just pointing out the obvious.

I don't think we are anywhere near the plateau on any curves. Our numbers aren't accurate, because even symptomatic people find it nearly impossible to get tested as far as I can tell. I mean, have you tried to figure out if you could get tested in your state?

[Ms. Toad]

7. According to today's curve,

the plateau would be April 24.

So you are correct, we are nowhere near the plateau. (I don't think anything I suggested implied we were.)

Think of the flatten the curve graphs - where we are headed right now is the long slow, climb uphill with a relatively constant number of new cases every day (when the multiplier is very close to 1), a period we will be in (if we're lucky) for more than a month before we hit the peak.)

If we're not lucky, when we hit that long slow steady climb people will get tired of isolating and we'll go back to exponential growth.

[JustFiveMoreMinutes]

33. Shouldn't caveats go first? n/t

[Ms. Toad]

37. Nope.

The caveats are there. I'm not going to encourage the unfortunate habit i see too much recently on DO of never reading past the headline.

[Ilsa]

4. I'm waiting for Florida's numbers to spike in a week or less

because of the spring breakers.

[33taw]

5. That is optimistic.

There are national and state curves. I think we are no where near the peak. New York may peak earlier Han other states, skewing these results.

[link:https://covid19.healthdata.org/projections|

[Ms. Toad]

11. Nothing I said suggested we were near the peak.

On the curve that fits from Feb 27 to today, the peak would be April 25. (Nearly a week later than the site you pointed me to.)

Remember, the entire goal was to flatten the curve. What that means is we will have a long, slow, steady hill to climb - in which the number of new cases will be relatively constant (instead of being based on the number of existing cases).

[JCMach1]

6. Death rate is probably the more important stat to track right now...

Just an example from Gov. Abbott in Texas...."blahblahblah... 18000 tests"

Sounds great, right!

5 minutes later: Emergency management judge in-charge of Dallas County, "there are significant clusters emerging in nursing and care facilities"

My note: There are 249 nursing homes in the DFW metroplex.


the math doesn't work...

[Ms. Toad]

14. The death rate will trail the new case rate by approximately 2 weeks.

I haven't checked, but that would suggest the death count is growing by about 1.4/day.

(Runs off to check . . . today's death count on worldometers is pretty low, suggesting to me it may not be complete. So taking the prior two days - 525/400 = 1.31. Expect the number of deaths to grow to around 1.5 by around Thursday {It's not an exact science, since people don't all die 2 weeks after infection})

[Lithos]

8. Operative word is tested

Cuomo did the right thing and increased tests in NY. Unfortunately, as this spreads into other areas, tests are not available, so these numbers will not count.

L-

[LisaL]

10. Yep.

Hard to get tested in many places. People are just told to stay home and assume they have it, if they have symptoms. But are those people counted as positive cases? Somehow I doubt it.

[Ms. Toad]

15. See China's curve on 2/12 & 2/13

As I suggested I would expect a similar jump over a short period when new tests are available.

I'd further expect it to be spread over a longer period of time, since China's was based on new diagnotic criteria. But it is coming at roughly the same point in the curve as China's brief burst of new cases. As soon as the new criteria settled in (two days), the rate of new case curve continued to slow.

One potential monkey wrench is that China had essentially one hot spot - and we have more than one (growing at somewhat different rates). But we'll see.

I'm just reporting what I see based on the numbers (as well as giving caveats - as I did with China when they started to bend their growth curve).

[Blue_true]

19. Like I pointed out last night, China took brutally drastic action,

going house to house to test people in the one hot spot, even testing people that had no symptoms to confirm they were ok. In the end, China had something like 65 million people that it was concerned about most, even as it clamped down on the rest of the country.

The virus has spread to every point in the USA, save really isolated places. We potentially have 300 million plus people exposed to this virus and we still have idiots meeting in big groups. I just feel that we are in unchartered waters, maybe New York, California and Washington State have gotten ahead of the virus, but we likely have new hotspots forming in big states like Florida, Texas, Georgia and controls from the state level in those places is bad.

[LisaL]

20. It's also likely they aren't exactly honest about how many people actually died.

They are claiming a pretty low number of dead people.

[Blue_true]

24. Yes, I noticed that, compared to Italy. nt

[Ms. Toad]

21. Just reporting observations,

parallels to what happened in other countries - and offering plenty of caveats (as I did with China) as to things that could make any predictions I make inaccurate).

I may be right - maybe not, but I've got a pretty good track record so far.

[Blue_true]

26. You points are very sound if big cities where Governors or Mayors

have acted slow the number of infections, that will slow the national rate. What people appear to be questioning is that hotspots are forming in a lot of new places, so the viral infection rates are accelerating rapidly in those places, look at the post where Missouri has seen a 600% increase, and no one knows where my state of Florida will head.

[Hoyt]

9. Have no issue with looking for hope in all this.

[sprinkleeninow]

29. And me as well. eom

[MyMission]

12. Here's an interesting video that explains it.

This is a very thoughtful data analysis from a YouTube math star

[Ms. Toad]

16. I've seen it before - it is very good!

Thanks for adding it here - especially for the people who are misreading my statement that the rate of growth of new cases is slowing means that the number of cases is about to plateau.

[Blue_true]

23. I don't think that people are challenging your statement about the virus

not plateauing. I think there is some concern that the data being reported isn't properly mirroring realities across the entire country. Hot spots are forming pretty much everywhere, a lot of those places are much smaller than LA, SF, NYC, Boston, Chicago, so as the big cities have success from lockdowns, isolation and social distances, infections are growing rapidly in less populated places - that very last point is why I believe people are questioning your analysis and projection, though if the big cities succeed at slowing the virus, you would be technically right as long as places like Florida and Texas don't become gigantic hotspots (at this time it looks like they are heading there).

[Ms. Toad]

25. Some are. It is those people I hope will watch the video. n/t

[Blue_true]

27. I think you are being a little too sensitive about people's critiques.

People are scared for themselves and their families. I wouldn't let it bother me if I were you, just keep running your analysis, like you wrote yesterday, time will tell.

[Ms. Toad]

28. It's not bothering me -

I just hope people will learn a bit about how these non-intuitive calculations work - and the video is a good tool for that.

[Lulu KC]

35. Very good--thank you n/t

[thucythucy]

13. I just now read here on DU that Missouri

has seen a 600% increase in cases over the past few days.

Plus--mega-churches in Florida and Louisiana had packed services this morning.

The measures taken in China were more draconian (and thus more effective) than anything we're doing here. Without a consistent national policy the surges in cases will simply shift from one part of the country to another, and then back again if shelter in place orders are lifted.

I'd prefer to be optimistic, but on the other hand I think the writing's on the wall. I hope I'm wrong, but I fear that I'm not.

In any case, best wishes to you and yours.

[Ms. Toad]

17. The growth in Missouri s included in the formula (since their numbers are in the mix).

So the rate of increase in new cases is slowing - even with the massive growth in Missouri

But you are correct that Florida and Missouriare wild cards that have the potential to tank what the rest of the country is doing, since they are not yet staying in place. I would be less worried about Louisiana (which does have a stay in place order) - except that there are already so many cases there.

[thucythucy]

30. Are these numbers also included?

https://www.democraticunderground.com/100213200196

That's 115 new cases in Tennessee in one nursing home alone. How many other nursing homes are similarly at risk? And notice that not only are residents infected, but also staff, who go home to their families, and who have presumably been going out into the community and no doubt infecting others. Who in turn infect others. None of which will become apparent for another week or so.

So Tennessee is then another "wild card" that might disrupt your more optimistic scenario.

As for Louisiana, we've just started to see cases that were infected a week or more ago. And many of the folks who attended Mardi Gras have now scattered back to their home states--raising the potential for an even greater spread.

As the saying goes--hope for the best, but prepare for the worst.

According to the experts our "best" case scenario is "only" one to two hundred thousand deaths. And the worst case scenarios are absolutely horrific.

One other factor to consider. Even if--and it's a big "if"--even if North America and western Europe are able to bring down the rate of increase, much of Latin America, the Caribbean, Africa and Southern and South West Asia are just beginning to experience their own outbreaks. Imagine the impact of 200 million cases in Vietnam, Cambodia, Burma, India, Sri Lanka, Indonesia. Imagine 50 million cases in Brazil (where the federal government seems every bit as inept and anti-science as our own).

The public health and economic devastation this will cause will dwarf anything we've seen to date. Who is going to do the planting in those regions, who's going to bring in the harvest? Food shortages will in turn lead to more vulnerable populations, which will exacerbate the impact of the pandemic. We saw this dynamic during the flu pandemic of 1919, where populations weakened by poor nutrition during the war were super vulnerable to the disease.

We are in for a very harsh time. Better to face up for it and prepare--which can only happen when the right wing know nothings--in the US, Britain, Brazil and elsewhere--are made to make way for people with some measure of competence and empathy.

I strongly suspect that it's way too early to claim anything close to success in slowing the spread, here or anywhere else.

Though I also very passionately hope I'm wrong.

[Ms. Toad]

41. You may, or may not, remember that I was one of the few on DU

Warning about this back in January, and being taken to the woodshed over it as an alarmist. I have been a lone voice in the wilderness warning about this in most of my social circles from then through early March. I'm hardly the one who needs to hear the message about how serious this -or that we should be preparing for the worst. That's literally all I've been preaching for the last two months.

I'm using the same spreadsheet I used to predict when the death toll from COVID 19 would surpass that of SARS, as well as one that is a bit more sophisticated, and reporting on the trends I see when they have continued long enough for me to be relatively confident they are not just data blips. (As I did with China) And adding in the things that I can envision that might disrupt the trend (as i did with China).

As for your specific question about Tennessee, it's not clear. There were 208 cases yesterday (the 29th on the site I use. There have been 98 since they started counting for the 30th. So they are either all included, or partially included, since there are fewer the 115 in the count for the 30th.

As for the rest of the world -these observations are specific to the US. It isn't that the rest of the world is irrelevant, it's just that it's not what I'm tracking, at the moment.

[Demovictory9]

22. good

[progree]

31. CDC has slightly different data ...

https://www.cdc.gov/coronavirus/2019-ncov/cases-updates/cases-in-us.html

Date     Cases     Factor   4-day Rolling average
18-Mar  010,442  1.484  1.480
19-Mar  015,219  1.457  1.455
20-Mar  018,747  1.232  1.460
21-Mar  024,583  1.311  1.371
22-Mar  033,404  1.359  1.340
23-Mar  044,183  1.323  1.306
24-Mar  054,453  1.232  1.306
25-Mar  068,440  1.257  1.293
26-Mar  085,356  1.247  1.265
27-Mar  103,321  1.210  1.237
28-Mar  122,653  1.187  1.225
29-Mar  142,047  1.158  1.201

CDC doesn't have March 29th yet (1243am ET 3/30), so I used Worldometer's for that one.

The factor is calculated the same as yours: number of cases divided by number of cases the previous day

The rolling average is simply the average of the date's factor and the previous 3 factors to smooth it. (My spreadsheet has data going back to March 8, I'm just showing the rows from March 18 onward)

I put a leading "0" on the cases so it would line up.

Even at a daily factor of 1.05 (meaning 5% increase per day, compounded), in a month the case load would be 1.05^30 = 4.32, a 4.3-fold increase

If it stayed at the last value of 1.158 it would be an 81.5 fold increase in a month. (Remembering that the 29th is from Worldometers while the 28th is from CDC so its not the perfect ratio to use).

[Ms. Toad]

36. It doesn't really matter which source, as long as you pick one

That consistently updated at the same time daily. They're all pulling from the same data sources.

The multiple day average is a good idea. Just remember you are using a trailing average, so it will lag the actual decline. I'm running that on mine, too, but I'm only using a three day average.

In a month, my formula has it dropping to about

But your basic point is correct, and is one at least a few seem to be missing - a decline in how fast the new cases are growing doesn't mean it is no longer growing at a rate proportional to the current number of cases (exponentially). It just means how quickly it is exploding explosion is slowing down.

[progree]

40. "In a month, my formula has it dropping to about"

please do tell what your formula is. I've been thinking of different regressions and curve-fitting.

I know that if the daily factor was constant, then a straight line linear regression would be a perfect fit to the log of the cases.

But since the factors are declining and the increase in the logs are slowing down, it is not ...

I just watched the video in #12 where they got into the logistic curve at 4:58 or thereabouts ...

Thanks!

[Ms. Toad]

42. It's a simple curve fitting polynomial.

Nothing sophisticted at all. I'm also running an exponential curve - but is nto as good a fit. It was not steep enough for a while, and is now too steep.

I just played with the polynomials in Excel until I found the degree I could reliably make fit - typically I have been able found one that fits with R^2 = .995+ (a pretty good fit). I adjust it daily as each new day's data comes in - I add a new point to the plot, replace the trend line and grab the new constants. It fits pretty well for about a week out (two days ago it was only off by 152; today about 200, if I recall correctly - genearlly within 1-2% of the next day's case total (for example). It has been a good ballpark general predictor for longer - for predicting within a day of when milestones farther than week or so out are likely to be hit.

Today is the first day the peak # of cases was wtihin the range I've been plotting (until today, both the polynomial and the exponential just kept climbing off the page).

[progree]

43. Thanks for the info. It's been a couple decades since the last time I did curve-fitting.

I'll have to dig up my notes, but what I might do when I get more time is:

Create a column of the log of the cases. E.g. Log(10,000 cases) = 4

If the daily percentage increases were constant for example, or random noisy around a constant, then a straight line would fit the log cases column, and Excel's simple linear regression capability would find the best fit line.

But since daily percentage increases are clearly trending down:

I'll fit a 2nd order polynomial to that log cases column, or a 3rd order polynomial, or whatever. Via Excel's linear regression functionality (which can be extended by some trick to fit polynomials or other types of curves)

I'll have to study the logistics curve too in that video and see if I can figure out how to get Excel to fit that.

Thanks again

[yaesu]

32. Oregons cases have been improving also which is where cases were first reported in the US

So certain regions will see improvement while others will get worse, its all about who's getting tested, when they were exposed & how they are containing it, a lot of variables.

[Warpy]

34. What will work once a thing like this escapes into

the general population is herd immunity, same as vaccines. Unfortunately, what will build herd immunity with this one is people who have caught it and survived it.

It's going to come at a high cost because there is no way yet to determine who will develop a severe illness and succumb and who will develop a light case and survive. However, in the absence of a vaccine, this is what will happen, the epidemic waning but a smaller number of cases still being reported until the vaccine is released--if the one in development works.

So as more people recover from it, the number of cases will diminish. It will continue to kill people until a vaccine comes out.

If antivaxers ever needed any proof of the superiority of vaccination, this disease should provide it to them.

[relayerbob]

38. Thank you, let's hope it continues that trend

[cstanleytech]

39. Assuming the math holds and there are no unexpected fuck ups due to Trump or his supporters

how soon then will we likely begin to see a decline?

[MyMission]

44. Fauci and other medical experts

Have now advised and advocated a lock down through April, so that's 1 more month, plus. I tend to trust their analysis and therefore believe=hope that we will see numbers plateau and decline by then. But there are several variables that effect spread, and they may decide at end of April to give it 2 more weeks.

Having said that, a YouTube psychic I follow, for entertainment purposes only, predicted that by mother's day we would see or be told about decline enough to return to "normal".

Mathematically, while we will see reductions in some areas, like Washington state, we will and have begun to see surges and clusters in new areas. Those will need to play themselves out. All the folks attending mega churches this weekend or next will have 2 week incubation, will spread it some and hopefully self isolate and new spread will decline.

Basically there's a predictive model, but it's a crapshoot!