New case growth continues to slow - updated 4/3

[Ms. Toad]

Last edited Fri Apr 3, 2020, 09:57 PM - Edit history (3)

12-day mostly downward trend in the growth of new cases (daily multiplier from prior day)- the two minor increases are bolded.

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
3/30 - 1.15344921
3/31 - 1.1506677084
4/1 - 1.140417971
4/2 - 1.138946898
4/3 - 1.131837616




(From original post) It's still growing exponentially - expect to see us blow past 200,000 tomorrow (likely before noon). The growth each day is just a smaller percentage of the current days cases. (In other words the number of new cases - as an absolute - is still going up each day, just not by as much)

I would not be surprised to see a signficant jump when Trump's mythical "everyone can be tested" tests materialize. But it is likely that the trend in the tested population is probably similar to the the trend in the population as a whole. See China's trend around the 2/12 & 2/13 jump.

(Added 4/2 - Still looks like the national peak is around 4/15 - simlar to other models)

[mucifer]

1. We don't have testing. I think a better clue as to how it's growing is

how fast the hospitals fill up.

[Jarqui]

2. +1

[JCMach1]

3. Exactly, there are no tests in Texas and many other places

[Ms. Toad]

8. If there are no tests in Texas,

Why did they report 760 new cases on the 31st, for a total of 3666?

Testing is absolutely limited - that's why I talk about trends, not absolute numbers. But diagnosed cases are likely to be proportional to the number of cases, since the most severe cases are generally being tested - and those will be a relatively consistent proportion of all cases.

[Texasgal]

16. There are tests in Texas.

The problem is that many people do not meet the criteria and labs are overwhelmed with processing.

It sucks, but there is active testing in Texas. FYI

[JCMach1]

21. Extremely, extremely limited... compare that per capita to what South Korea did...

[Ms. Toad]

24. It bad all over in the US - not just Texas. n/t

[Texasgal]

26. I don't disagree with that.

Yes... it's limited. But you said there were no tests in Texas and that in incorrect.

Do we need more tests? Hell yes! Do we need faster results? Hell yes!

In Austin, testing sites have changed labs to get quicker results. I hope this goes state wide.

[Ms. Toad]

5. Not really.

How fast hospitals fill up is only indirectly related to how many new cases there are - since people are not hospitalized and released on the same day. They enter and then leave anytime from a day or two - to 3 weeks (or more) later. So how quickly the hospitals fill up isn't a measure of growth in new cases - and it is the trend in new cases that will tell us whether we have flattened the curve.

I find it intersting that I was being trashed when I predicted this back in January by accurately predicting China's rapid increase in daily cases - and thought I was being "chicken little." And now people who thought my early trend tracking and predictions were wildly exaggerated are now telling telling me I don't know what I'm talking about because I am observing a downward trend in growth of new cases (and am suggesting the trend is no longer a rapidly increasing growth rate).

My methodology hasn't changed from mid-January (aside from adding one new curve). I'm relatively confident in the trends I'm reporting (not absolute numbers - but trends). They haven't been wrong yet.

[pstokely]

38. what about those not sick enough for hospitalization?

they may not test those

[NRaleighLiberal]

4. hi, my fellow data pal

I am still keeping my big OP current - making lots of edits. Noted that the weekend saw more of a lessening - wonder if it is a "weekend" effect - haven't gone back to see if the confirmed cases tend to reduce on Sat and Sun.

[Ms. Toad]

6. One of the two very slight upticks in growth was a Monday -

which makes sense. The other was on a Thursday.

It's not a linear drop - so there may be big drops over the weekend and less during the week. I haven't looked at at that granular a level.

[NRaleighLiberal]

7. I wonder how long 1.15 factor will hold - so many competing variables

amount of testing, etc - surprised at how predictive it has been, really

[Ms. Toad]

10. My polynomial curve -

that matches the current curve extremely closely, has little over 2-week decline in case growth to the peak (going down about about .01 to about .008/day).

[dweller]

9. i watch 2 numbers

total deaths and total recovered
the further these 2 grow apart with recovered getting larger
the better it seems to be

just my 2 cents
✌🏼

[Ms. Toad]

11. I'm not sure that captures when the growth in new cases starts slowing

since that lags by at least 2 weeks.

[herding cats]

12. Nationally, yes. My state (Texas) keeps ticking upward on the curve.

Here testing is extremely rare due to shortages, and results are taking up to two weeks to come in, so the numbers are extremely depressed and slowed. Even still they're consistently ticking up.

I'm on a stay at home order in my county. It's so lax most service industries still are expected to work as "essential" businesses. Literally, the housekeeper and the lawn person are considered essential here, among many other non-essential businesses. Which means they don't qualify for the expanded unemployment and have to work still to pay their bills. We're forcing way too many non-essential service workers to keep working to be able to eat in states like this. There will be a backlash.

But, we can get drive thru pitchers of margaritas, so at least there's that. 😑

[DrToast]

13. Yes, it doesn't make sense to consider it as one outbreak.

[herding cats]

14. Exactly.

Each state sets their own standards right now. Which means as this spreads outward numbers will most likely change. It's the nature of or states rights system.

[Ms. Toad]

30. There are going to be places ahead and behind the curve

But there is still a very smooth national curve, at this point.

[Ms. Toad]

15. I don't have a source for day-to-day data in Texas

so I can't compare it with the rest of the US. The US still has a bigger number of new cases than the prior day - it's just that it's not as much bigger each day - the hill is leveling off.

Enjoy those drive thu pitchers of margaritas - just don't drink them until you're in a place you plan to stay for a few hours.

[herding cats]

18. I definitely agree the hill, Nationally, is leveling off.

I've been tracking it as well. My suspicion is it's due to truly solid social distancing in our most heavily hit epicenters so far.

San Francisco is a model to follow, and NY state is working so hard to fight this virus. They make me so proud to see them do their best when faced with this novel virus.

Other states are worrying me with their approaches, but time will still tell if heat and humidity slow this virus down. Louisiana is a concern on that front, but the massive exposure Mardi Gras created is an exception, perhaps? If heat and humidity slows it down turns out to be true, we're on a good trajectory. Possibly a very good one, even.

Sadly, I stopped drinking before Texas went to its drive thru COVID-19 alcohol laws. I was drinking too much for a bit after Trump was elected. I'm rethinking that decision now.

Your math is sound. I don't want you to think I'm not respecting your skills. I agree with your analysis at this point.

[Ms. Toad]

19. I was just curious to see what the Texas curve looked like -

But I can't find retrospective daily Texas numbers. I didn't take your comments as challenging the trends I was seeing.

[JCMach1]

22. Texas is going to be all over the map due to the Governor's inaction, but positive action by some

local officials...

[herding cats]

25. I've been following here:

https://coronavirus.1point3acres.com/en

We're growing by about 100 actually tested per day. Which is a fraction of the now considered presumptive positive cases being reported. That's the issue here. We have an extremely limited number of available public tests. I've viewed our city and county meetings validating such. We're in a unique position here in Texas where I've even been witnessing local to me red counties saying they can't get the needed testing just two day ago. As they said, they're a low priority.

Pencil the numbers down and keep in mind testing is rare and results are up to 2 weeks to show. It's a peculiar situation.

[Roland99]

17. If I understand your numbers correctly...

3/31 - 1.150667708

Would be 3/30 total new cases multiplied by 1.150667708? Or is it a multiplier of total cases?

In any case, I’d expect that number to drop somewhat as time goes on. It’s a bit misleading, imo, to suggest a slowing

For example. If there were 18k new cases on 3/29, we’d have 20,672 new cases on 3/30. Then we’d have 23,890 on 3/31

A slightly smaller % multiplied by a larger number will still be a much larger number

We can’t expect 30% or 20% growth every day.

We’d need to see that multiplier drop much more quickly. But as testing expands, we’ll likely see it start to go up

And the graph here doesn’t seem to show and flattening of that upward slope

https://coronavirus.1point3acres.com/

[ProfessorGAC]

34. Depends On Data Point, Roland

Using, for instance, 16 to 20 days of data, small differences can still exhibit a valid trend. That said, unless you've got at least 30 days, any one day well of the fit line spoils the analysis.
I agree it would be useless to only use 5 days, or something.
My biggest concern about the data shown is that differences are so tiny day to day. And 2 of those 9 days, the number went up.
Using 9 decimal places starts to fractionalize people. Can't have 1% of a person, or be 1% infected.

[Roland99]

35. %s do work better than pure numbers

Look at the stock market. A 1,000pt drop in the Dow sounds ominous but, even at these levels, it's less than 5% and would probably not even make the top 50 all-time % losses for a day.

Or take a salary increase. Someone making $25k/yr gets a $1,000/yr raise. That's much more beneficial than someone making, say, $150k/yr and getting the same $1,000/yr raise.


Seeing these cases go up 20k...23k/day seem ominous but what are the underlying %s.

[ProfessorGAC]

36. Same Page

The data set of 9 is too small to predict a trend, when 22.2% of them actually show daily increases.
When we've got 30 days, we can revisit.

[Roland99]

37. US seems to really be stair-stepping in terms of new daily cases

See 2nd graph here:
https://www.worldometers.info/coronavirus/country/us/

Wonder if that's a reporting effect?

[ProfessorGAC]

39. By Coincidence...

...I looked at that very chart yesterday afternoon.
We do have an issue with reporting. Some of it is as simple as a day has 24 hours. 12 midnight to 11:59 pm.
For convenience, let's just do 1000 cases a day.
One day the data gets reported for the day at exactly midnight. 1000.
Next 2 days it's reported at 11:50 then 11:45pm. The second day reports 7 fewer cases. The third day would be 3 more short, but the 7 from the day before end up there. So the hard number goes up.
Fourth day it's reported right at midnight. The entire 10 left over from the day before & that day is 1010.
But, if we have enough days of data, we can do a trend line and look for meaningful slopes up or down.
Right now, we're all working with squishy data, in reporting block over only 15 or 20 days in true crisis mode.
To reply see what's going on, probably need Kore time.
BTW: if you look at the first chart, notice that the last 8 to 10 points are a straight line. Limited data points, but that suggests constant daily growth. Most likely a linear increase in testing. Again, establishing the squishy data thing.

[Roland99]

40. " the last 8 to 10 points are a straight line"

and that's not a good thing. It's not bending into a curve yet.

[ProfessorGAC]

41. Dead On

It suggests a linear increase in testing. They're testing a higher number every day. But that increase is not exponential. So the more tests, the more positives.
That said, if distancing was doing nothing we could expect the % of the new tests to be rising faster than the rate of new testing. Test 5k. Get 50% hits. Test 6k, 50%. We see an increase of 500 cases.
But if nothing helped, the second day would be 55% of 6k, or 3300. Then 60% 7k or 4200.
Those points suggest the number of infections is static to the increase in tests. The (made up) infection rate stays 50%.

It's only 8 days, but it at least suggests that distancing is helping something.

[la-trucker]

20. You forget that the red states where there is no forced isolation

have not kicked in. There are still states that have <500 cases -- when they grow, the daily growth rates would spike again.

[Ms. Toad]

23. I'm not forgetting anything.

I'm tracking the actual numbers, which include both states that have imposed social distancing - and those that haven't. So they're in there - and influencing the trend I'm reporting. I'm running a best fit curve on data that includes the data from every single state - as I did with China (which included both severe lock-down - and less severe, and lock-downs on different timing).

And not every state will necessarily exceed 500 - it will depend on where the virus is, how densely populated the state is, how much they move around.

What is going on right now is that the rate of increase in cases each day is declining. (in other words there is still an increase in the absolute number of new confirmed cases every day - it has just been a smaller increase each day for the last 9 (aside from two which were each within 1% of the prior day's increase).

[JCMach1]

27. Just one point... notice drop on 3/29 then a rise again. Weather affects outdoor testing

I know at that date, there were 0 drive through tests conducted in DFW.


It will take a trend line of several weeks to clear out the noise. Factors that you might not be able to control for:

-Affect of weather on testing
-number/availability of tests in a particular area
-correct reporting of cases... (For example Tarrant Co. in TX. now includes 'presumed cases' and dropped all of those on today's data.
-Cases that just don't/won't test (this will eventually be a very large number)

It's me, but I am going to fall on the pessimist side of the equation hoping FFS I am wrong.

[Ms. Toad]

46. My trend line is 6 weeks long.

And that was accurate enough to accurately detect China's similar slow down in rate of new cases.

[Loki Liesmith]

28. Testing has plateaued and trended a bit downward tho

So the real number of new cases is somewhat obscured to us.

Also a weird weekend effect where smaller numbers are reported on weekends.

[Ms. Toad]

47. Only one of the two blips in this trend is a weekend day. n/t

[Loki Liesmith]

49. That is clear yes

[Demonaut]

29. these are minor anomalies, we saw something in China like this too, this fire has a ways to go

[Ms. Toad]

31. Correct - we saw something like this in China.

BUT - They were not minor anomalies in China - which is part of my point.

When we saw this turndown in growth of new cases in China, the tide had turned. It does NOT mean the fire is out - a decrease in growth rate when you start at a rate as high s 1.5 simply means the hill is getting flatter - it doesn't mean you are over the top of the hill.

My curve matches that of all the experts I've heard - in terms of when it will reach the maximum number of new cases - ~April 15. (After that, the number of new cases will get smaller each day.)

Until thereabouts, the absolute number of new cases will continue to increase every day - As long as the multiplier is larger than 1.0, each day will continue to have more new cases than the last. What the trend shows is merely that it is not ballooning as quickly as it was..

[NewsCenter28]

32. If OP is accurate, this news would be tremendous for the KGOP

Unfortunately. I believe you said that we'll peak on April 15th, correct, Ms. Toad? I hate to be partisan and crass here, but your projections would set Trump up nicely to re-open America on April 30th as he promised and act like the conquering hero all the way to November and make Coronavirus but a distant memory by November 3, 2020, with how fast the news cycle moves these days. A re-opening of America on April 30, 2020, or May 1, 2020, etc., would set the economy and stock market up for a boom just as it is time to start voting in November.

I don't understand why we can't have the Olympics and conventions then if we're just 2 weeks away from this falling off and essentially disappearing soon, like in China?

Interesting scenario. Honestly, I'm just trying to learn here and not trying to be overly presumptuous but wouldn't your modeling and projections indicate that the US won't see anywhere close to 100 K deaths, let alone 200 K, and that the projections are way overblown? Perhaps, an effort by the Toad-in-Chief to make it look like he conquered a far more daunting enemy than actually existed to aid in his re-election efforts. Again, I understand that would be a good thing. Better to err on the side of caution with something so potentially dangerous.

[Ms. Toad]

33. The data is correct - and it shows a current slow down in the rate of new cases.

Whether that trend is predictive of the future is what is less certain. In China it was, and it has been in the US for general targets (e.g. whe are we going to hit X cases), and it has been a very close predictor of the next day's new cases.

So the peak currently looks to be April 15 - and that has remained my predicted peak for about 3 days (and it is consistent with the peak predicted by experts).

My guess is that once Trump realized he couldn't just talk this thing away, and got his experts to predict relatively safe endpoints and maximum so that reality would match - or be slightly better than what he predicted yesterday - for precisely the reason you suggest. An article I read last night suggested that he was told by his advisors that he wouldn't own the economy because there is a global pandemic, but that h would own the deaths if he did nothing. So he did, as he always does: he pretended he had never said anything at all before on this issue and shifted course.

I haven't been running a death curve - so is hard to tell how the two relate. I just pulled one together - it appears to be running about 5 days behind (which seems a bit short to me). If that is accurate (and the curves hold) - my "guestimate" is that total death toll would be between 170,000 and 200,000. (That is based on forcing the curve I can see now in the new cases trend onto the death curve with a delay of 5 days (for the 170,000) and 200,000 (the current unaltered prediction - just where the formula takes me). I have no idea how realistic these are - since I had to make a number of assumptions to get there. That said, it's similar to the numbers predicted by Fauci.

I think everything has been cancelled through July because (1) models aren't perfect - and many of the things cancelled involve lots of advance expenses that cannot be recouped if it has to be cancelled at the last minute and (2) It's our turn in the spotlight now - there are other countries who are behind us - they may be peaking in (or closer to) July.

[greenjar_01]

43. The two of you!

[Ms. Toad]

42. Bump for the 4/1 edit. n/t

[Alex4Martinez]

44. This page shows higher numbers of deaths for US:

https://ncov2019.live/data/usa

[Ms. Toad]

45. What I am noticing is not that new cases are lower -

but that the rate of increase in new cases is slowing.

New cases were growing by as much as 1.5 x the prior day's cases. They are now growing at the rate of 1.14 times the prior days cases.

There are still more new cases every day - just not as many more new cases each day. The difference between accelerating by flooring the gas pedal and gently pushing the gas pedal. We had been flooring it - we're moving closer to gently pushing it. Still going faster each second - but instead of going from 0 to 100 in 6 seconds, we're now going from 0 to 100 in a minute.

Deaths always lag new cases. Using my formula they appear to be about 5 days behind.

Today's increase in the growth of deaths declined. The multiplier from yesterday's deaths was 1.26. The prior day was 1.29. I never trust a one-day decline - but it would be consistent for the increase in the growth of deaths to start to decline.

As with new cases - I'm not suggesting that there are fewer deaths - just that the lag time is about right for deaths to be easing off the accelerator a bit.


ETA: The deaths on the site you referenced is identical to the number on the site I'm using - at least for today.

[Alex4Martinez]

48. Yes, I see that and have been checking the rates of increase for my state.

Of course, it's all dependent on the number of tests administered and the manner of selection.

Still, rates of 1.37 day over day were dropping slowly over time and then I found new sources and have been comparing projections for various states.

Peak numbers come at different times for different states. Here in California we can expect peak demand for beds to come at the end of April but for NY and NJ it's just two weeks out.

Other states, in that sense, are our future.

My go to links are these:

https://covidactnow.org/ Projections for each state, individual graphs

https://covid19.healthdata.org/projections Also, graphs for each state

https://ncov2019.live/data/usa

https://coronavirus.jhu.edu/map.html

https://www.ft.com/coronavirus-latest

https://bnonews.com/index.php/2020/03/tracking-coronavirus-u-s-real-time-visualizations/

[Ms. Toad]

50. I was looking for that first link earlier today & couldn't find it.

My peak new cases nationally (as of today) is April 14 (two days shy of the peak resources map). So it seems to be relatively close.

[Ms. Toad]

51. Kick for 4/3 update - rate of growth is still slowing!!

Last edited Fri Apr 3, 2020, 09:57 PM - Edit history (1)