Coronavirus 2019-nCoV

[SB Shock]

BTW -- You asked a very interesting question that I didn't think of when looking at this data.

Today's additional data points jumped R² over .70 at 0.70258.

Has anybody on here ever seen real (non-academic) data sets have that high of an R²?! I haven't. There are almost always too many variables to get two separate data sets to yield that type of determinant. That's eye popping.

Source to data set?

[RoyalShock]

Yep. I understand they could. Typhoid Mary is a great example of proper use of such an authority.

This person in Reno is an ass. Somewhere above voluntary quarantine, but less than prison is probably the correct response.

My understanding (and this is not public information) is that the situation in Reno Co was contained within one family,and as far as is known, was not a case where the general public was exposed. It was still an incredibly poor decision the person made.

[SB Shock]

Harper county just picked up their first case - in a rest home in Attica.

[Kung Wu]

Source to data set?

https://covidtracking.com/api/v1/states/daily.json

But I have a program that filters the Kansas data, so let me just run it and grab the raw data for you.

[Kung Wu]

Harper county just picked up their first case - in a rest home in Attica.

I cut off on 3/27 because there is no hospitalization data prior to that, and that's what the Kungculator is currently looking at.

Column 3 and 4 are the columns of interest.

The columns are:

Date, Total cases, Increase in cases, Increase in tests, Increase in hospitalization

20200429	3738	247	1368	11
20200428	3491	163	923	8
20200427	3328	154	940	11
20200426	3174	118	1028	11
20200425	3056	279	1611	17
20200424	2777	295	2270	15
20200423	2482	271	1115	10
20200422	2211	186	1102	13
20200421	2025	39	340	14
20200420	1986	137	631	15
20200419	1849	59	454	7
20200418	1790	85	775	8
20200417	1705	117	779	16
20200416	1588	94	1454	17
20200415	1494	68	521	15
20200414	1426	50	283	18
20200413	1376	39	611	11
20200412	1337	69	910	5
20200411	1268	102	929	19
20200410	1166	60	639	11
20200409	1106	60	592	40
20200408	1046	146	669	0
20200407	900	55	430	25
20200406	845	98	861	15
20200405	747	49	645	11
20200404	698	78	504	21
20200403	620	68	463	13
20200402	552	70	718	24
20200401	482	54	469	35
20200331	428	60	502	13
20200330	368	49	409	11
20200329	319	58	581	28
20200328	261	59	501	0
20200327	202	34	394	27

[SB Shock]

I cut off on 3/27 because there is no hospitalization data prior to that, and that's what the Kungculator is currently looking at.

Column 3 and 4 are the columns of interest.

The columns are:

Date, Total cases, Increase in cases, Increase in tests, Increase in hospitalization

20200429	3738	247	1368	11
20200428	3491	163	923	8
20200427	3328	154	940	11
20200426	3174	118	1028	11
20200425	3056	279	1611	17
20200424	2777	295	2270	15
20200423	2482	271	1115	10
20200422	2211	186	1102	13
20200421	2025	39	340	14
20200420	1986	137	631	15
20200419	1849	59	454	7
20200418	1790	85	775	8
20200417	1705	117	779	16
20200416	1588	94	1454	17
20200415	1494	68	521	15
20200414	1426	50	283	18
20200413	1376	39	611	11
20200412	1337	69	910	5
20200411	1268	102	929	19
20200410	1166	60	639	11
20200409	1106	60	592	40
20200408	1046	146	669	0
20200407	900	55	430	25
20200406	845	98	861	15
20200405	747	49	645	11
20200404	698	78	504	21
20200403	620	68	463	13
20200402	552	70	718	24
20200401	482	54	469	35
20200331	428	60	502	13
20200330	368	49	409	11
20200329	319	58	581	28
20200328	261	59	501	0
20200327	202	34	394	27

So you are plotting columns 3 vs 4 and getting R^2 of 0.70?

[Kung Wu]

So you are plotting columns 3 vs 4 and getting R^2 of 0.70?

I used this online calculator, but I think it's suspect because that R² seems very high.

[wufan]

So, I like to drive fast cars. Does that mean that if I drove recklessly and killed someone while showing off, that would just make me a dangerous asshole who did not deserve jail time? Just asking how far we go with that label.

That’s manslaughter.

[SB Shock]

I used this online calculator, but I think it's suspect because that R² seems very high.

I have plotted columns 3 vs 4, I do get a R^2 = 0.7026. So nothing wrong with what you are doing. Now I'm doing some thinking of what plotting those two variable actually mean?.

Because it is true they are increasing testing. And you would expect your aggregate numbers to climb because you are testing more. So that is valid thing to be looking at.

Since they want robust testing, the requirements is not to show that you have declining in raw aggregate of cases (that would lead people to test less, not more), so what that government asked is you to show that your positive % is actually declining as you ramp of testing which means you are being more thorough in your testing of the population.

So what the R^2 = 0.70 is actually telling is since it is not 1.0, you need to ask what is causing the variability? You variability is being affected by % of positive and negative cases. Higher negatives can be drive your variability (which is good), or higher positive rates can drive your variability (which is bad)

Here is the average for the 4 weeks of data (week 4 includes 8th day)

Week 1: 511 test per day
Week 2: 595 test per day
Week 3: 764 test per day
Week 4: 1322 test per day

Here is the average of for the % positives

Week 1: 10.7% tested positive
Week 2: 13.3%
Week 3: 9.0%
Week 4: 16.5%

Week 3 is good trend, but week 4 is not the trend you want. You want your cases % to drop even as you ramp up testing.

In week 4, if your increases in cases was just due to testing (assuming 11% positive rate - this is based on the previous 3 week average), then the cases would be 145, but the actual average was 218.

What has happened in the last week? Ford and Seward counties - 1175 positives, 1447 negatives (45% positive rate).


trends.JPG

[Aargh]

In an odd turn of events, Kansas seems to have its first herd immunity. It's the prisoners at Lansing.

Guards and staff had to work without PPE. 75 of the staff tested positive, so they tested all the prisoners. 75% tested positive. That wasn't the antibody test, that was the presence of virus test.

One other interesting observation. 50% of the prisoners who tested positive had no symptoms.

[wufan]

In an odd turn of events, Kansas seems to have its first herd immunity. It's the prisoners at Lansing.

Guards and staff had to work without PPE. 75 of the staff tested positive, so they tested all the prisoners. 75% tested positive. That wasn't the antibody test, that was the presence of virus test.

One other interesting observation. 50% of the prisoners who tested positive had no symptoms.

I’m guessing we will have a 50% asymptomatic population in the immunocompetent population.

[Aargh]

US CV deaths March 31: at least 3,800
US CV deaths April 30: at least 63,000

That's 59,000 deaths in April.

CA and NY are heavy population areas and they are both past their peaks. With the gradual reopening, the back side of the peak is likely to have a shallow slope. If the estimate of 74,000 deaths by the end of August is accurate, that would be incredibly encouraging compared to what we've seen this month.

[wufan]

US CV deaths March 31: at least 3,800
US CV deaths April 30: at least 63,000

That's 59,000 deaths in April.

CA and NY are heavy population areas and they are both past their peaks. With the gradual reopening, the back side of the peak is likely to have a shallow slope. If the estimate of 74,000 deaths by the end of August is accurate, that would be incredibly encouraging compared to what we've seen this month.

There’s been a lot of mortality prediction models. One of the models changes as more data comes in. Others were more of a prediction at point A and were never remodeled. Some states changed the requirement for what’s a Covid death. I’m curious through all of this if anyone has tracked the best fit/worst fit models?

[wsushox1]

In an odd turn of events, Kansas seems to have its first herd immunity. It's the prisoners at Lansing.

Guards and staff had to work without PPE. 75 of the staff tested positive, so they tested all the prisoners. 75% tested positive. That wasn't the antibody test, that was the presence of virus test.

One other interesting observation. 50% of the prisoners who tested positive had no symptoms.

These are similar numbers to prison populations in North Carolina.

[wsushox1]

In regard to the R2 discussions,

If the R0 of the virus is constant, you would expect a fairly high correlation between tests and confirmed positives - correct? But as SB has stated, % positive of available tests is probably the best metric to look at.