Toggle sidebar

Extrapolating the Covid-19 Data

S

Soaring Eagle 74

Freshman
Jan 4, 2008
22,888
69
0
If you're not into data analysis, don't waste your time reading this. I'm not a biomedical analyst.

I loaded-up the data on reported Covid-19 cases by date and tried to model it. The result was:

Reported Cases = (Days Since Inception) x 1.223^0.738

That gave me an R^2 correlation coefficient of 0.9988 (1.00 is perfect)

That suggests 927,000 reported cases and 22,243 deaths by Easter Sunday and 13,431,819 reported cases and 322,364 deaths by the end of April.

I certainly hope these numbers are way too high. There are no adjustments for better isolation and hygiene practices. Also, with each passing day there are fewer people in the pool of the uninfected, so fewer people to get infected.

I model static systems as opposed to dynamic systems like Covid-19, so there's a good chance I'm way off. However, I haven't seen calendar-based forecasts so I tried to see what the data might suggest.

Data source: https://covid.bio/growth

Keep safe.
 
Last edited:
P

Pospecteer

All-Conference
Dec 8, 2006
36,507
3,172
113
Soaring Eagle 74 said:
If you're not into data analysis, don't waste your time reading this. I'm not a biomedical analyst.

I loaded-up the data on reported Covid-19 cases by date and tried to model it. The result was:

Reported Cases = (Days Since Inception) x 1.223^0.738

That gave me an R^2 correlation coefficient of 0.9988 (1.00 is perfect)

That suggests 927,000 reported cases and 22,243 deaths by Easter Sunday and 13,431,819 reported cases and 322,364 deaths by the end of April.

I certainly hope these numbers are way too high. there are no adjustments for better isolation and hygiene practices. Also, with each passing day there are fewer people in the pool of the uninfected, so fewer people to get infected.

I model static systems as opposed to dynamic systems like Covid-19, so there's a good chance I'm way off. However, I haven't seen calendar-based forecasts so I tried to see what the data might suggest.

Data source: https://covid.bio/growth

Keep safe.
Click to expand...

This is a much better way to spend your time....as opposed to the typical Trump bashing you normally do.
 
M

MikeRafone

Freshman
Oct 5, 2011
4,238
53
0
I was doing Monte Carlo simulations for politicians in the mid-80's. If you think this stuff is fun now, try to imagine what it was like then? Electronic calculators were about as high tech as it got as was loading the info into to the Commodore/Tandy boat anchors, with that miracle of the age, the floppy disk. Miss a slash and 20 hours of data/code was gone.It was real state of the art stuff. If COV keeps chugging along, I know a few fine jumping cliffs/bridges across the country. Don't limit yourself to the Star City Bridge. Dead is dead, even if the Church ladies are arguing about "Don't she look natural?"

My youngest bullheaded daughter loves Morgantown. I've been trying to talk her into moving to Ft. Collins, CO with her fun Aunt to dodge the greater NYC bomb squad of hormones that's going to be rolling into Motown with the young students next August and their last chance for a fine romance or two in their home towns. They then hit town spreading it across Morgantown this fall.

The way I look at,she can get her WVU degree by mail or online and Dad can hit some of the legendary trout steams throughout the Colorado region. It's a win-win situation.
 
M

MikeRafone

Freshman
Oct 5, 2011
4,238
53
0
BTW, viruses of this type are worse the second year around. Second grade, I got what was called the "German Measles" in those days. In a split 3rd and 4th grade that started with 34 kids, only six were healthy going into April. They closed school for two weeks to wait it out. We had a few more straggle in towards May, but the school board and the Health Dept. were as mystified by it as everyone else was.
 
W

WVU82_rivals

Senior
May 29, 2001
199,091
693
0
 
S

Soaring Eagle 74

Freshman
Jan 4, 2008
22,888
69
0
WVU82 said:
Click to expand...

What do you think this graph shows?

I wonder if the decline in deaths is a reflection of the home isolation and reduced traffic accidents. One might conclude from this graph that the corona virus is good for US death rates. I wouldn’t. Correlation isn’t causation.

The source of the graph is Evviva brands, which appears to be a marketing company based in California. They don’t reference a source for the data points shown, so a bit of skepticism might be warranted.

evvivabrands.com/covid-19/different-model

Here’s a quote from the web page:
“Fewer highway miles. Fewer traffic crossings. Fewer hours in hazardous environments. More sleep. Less physical stress. Cleaner air. More hours available for parental care. Who knows how many factors contributed to this decline? What do we need to do to capture these death-defying insights and codify them into policy?”
 
V

va87eer

Freshman
Jan 16, 2006
2,564
70
48
Soaring Eagle 74 said:
If you're not into data analysis, don't waste your time reading this. I'm not a biomedical analyst.

I loaded-up the data on reported Covid-19 cases by date and tried to model it. The result was:

Reported Cases = (Days Since Inception) x 1.223^0.738

That gave me an R^2 correlation coefficient of 0.9988 (1.00 is perfect)

That suggests 927,000 reported cases and 22,243 deaths by Easter Sunday and 13,431,819 reported cases and 322,364 deaths by the end of April.

I certainly hope these numbers are way too high. There are no adjustments for better isolation and hygiene practices. Also, with each passing day there are fewer people in the pool of the uninfected, so fewer people to get infected.

I model static systems as opposed to dynamic systems like Covid-19, so there's a good chance I'm way off. However, I haven't seen calendar-based forecasts so I tried to see what the data might suggest.

Data source: https://covid.bio/growth

Keep safe.
Click to expand...
Soaring Eagle 74 said:
If you're not into data analysis, don't waste your time reading this. I'm not a biomedical analyst.

I loaded-up the data on reported Covid-19 cases by date and tried to model it. The result was:

Reported Cases = (Days Since Inception) x 1.223^0.738

That gave me an R^2 correlation coefficient of 0.9988 (1.00 is perfect)

That suggests 927,000 reported cases and 22,243 deaths by Easter Sunday and 13,431,819 reported cases and 322,364 deaths by the end of April.

I certainly hope these numbers are way too high. There are no adjustments for better isolation and hygiene practices. Also, with each passing day there are fewer people in the pool of the uninfected, so fewer people to get infected.

I model static systems as opposed to dynamic systems like Covid-19, so there's a good chance I'm way off. However, I haven't seen calendar-based forecasts so I tried to see what the data might suggest.

Data source: https://covid.bio/growth

Keep safe.
Click to expand...

I've been doing something similar since early March.. If your model extrapolates the current curve (based on the latest data), you're likely to shoot very much higher than the actual figures. I was doing that for the first week and noticed that in a few days the calculated values were already diverging high from the actuals by quite a bit. If you also factor in the way the curve is changing from day to day you'll get a much closer answer. In other words, don't just add a new data point. Add the new data point, refit the curve, and see how the curve fits are changing over time. It could be that your simulation is already doing that. At some day (hopefully very soon), the curve in new cases should take a sharp turn toward flattening due to the isolation activities and the fatality graphs will likely do that a week or two later.

Stay safe!!
 
S

Soaring Eagle 74

Freshman
Jan 4, 2008
22,888
69
0
va87eer said:
I've been doing something similar since early March.. If your model extrapolates the current curve (based on the latest data), you're likely to shoot very much higher than the actual figures. I was doing that for the first week and noticed that in a few days the calculated values were already diverging high from the actuals by quite a bit. If you also factor in the way the curve is changing from day to day you'll get a much closer answer. In other words, don't just add a new data point. Add the new data point, refit the curve, and see how the curve fits are changing over time. It could be that your simulation is already doing that. At some day (hopefully very soon), the curve in new cases should take a sharp turn toward flattening due to the isolation activities and the fatality graphs will likely do that a week or two later.

Stay safe!!
Click to expand...

I think there are so many variables that I haven’t considered the good correlation may be a function of self-cancelling errors. For example, I suspect there could be twice as many cases as actually reported and a lot more deaths as well.
 
V

va87eer

Freshman
Jan 16, 2006
2,564
70
48
Soaring Eagle 74 said:
I think there are so many variables that I haven’t considered the good correlation may be a function of self-cancelling errors. For example, I suspect there could be twice as many cases as actually reported and a lot more deaths as well.
Click to expand...

It was surprising to me that in the early days for almost every country the increases in cases per day was very consistent at about 33% per day +/-5%. With so many variables in play like social norms, population density, testing rates and a hundred others, almost all countries have that rate once they go over 100 cases until social distancing takes effect. It's a pretty fascinating subject.

In the attached file this guy explains some simple models that he built to test the sensitivity to different variables. He's not a specialist in the field and is looking only at simple statistical models, but it's still interesting.
 
O

op2

All-Conference
Mar 16, 2014
11,690
1,301
103
Soaring Eagle 74 said:
What do you think this graph shows?

I wonder if the decline in deaths is a reflection of the home isolation and reduced traffic accidents. One might conclude from this graph that the corona virus is good for US death rates. I wouldn’t. Correlation isn’t causation.

The source of the graph is Evviva brands, which appears to be a marketing company based in California. They don’t reference a source for the data points shown, so a bit of skepticism might be warranted.

evvivabrands.com/covid-19/different-model

Here’s a quote from the web page:
“Fewer highway miles. Fewer traffic crossings. Fewer hours in hazardous environments. More sleep. Less physical stress. Cleaner air. More hours available for parental care. Who knows how many factors contributed to this decline? What do we need to do to capture these death-defying insights and codify them into policy?”
Click to expand...

That quote implies we'd all be better off if this continued indefinitely. First of all, tons of people are losing their jobs and the money that comes along with it. The government will bail a lot of them out but then that just adds more that has to be paid eventually by taxpayers.

On top of that, I bet there is tons more anxiety and stress now than before, which has harmful effects but in the long term and not the short term.

All that said, it wouldn't surprise me if there weren't some positives from all this in the end. People are re-assessing what is important in life and what is not.
 
S

Shirley Knott

Redshirt
May 26, 2017
12,831
0
0
Imagine a SWAG thread......Scientific Wild Assed Guess.....Who knew souring aggle was a scientist......stay safe !
 
You must log in or register to reply here.
Top Bottom