COVID-19 Statistical Prediction Model

[BoilerBulldog]

Not sure If anyone has come across this guy’s model for predicting how COVID will spread and then slow down. Apparently it’s been tracking very closely so far. If his model is correct, US will peak on 3/28 and we will end up with around 1400 deaths on 180,000 infections. Anyone with a stronger stat background than me want to take a look and give me your thoughts? This is a far more positive outlook than many are predicting.

https://dispensationalpublishing.com/covid-19-scientific-prediction-model-march-23/

 

[HomelessHarvey97]

AMARCOTT my boy get on it!

 

[Crazy8]

And @JoeZap.

 

[Tkal05]

Boy. Who wouldn't sign up for that right now? Too lazy to open link - does his modeling factor in existing cases that haven’t been diagnosed - and their impending spread?

 

[BoilerBulldog]

Boy. Who wouldn't sign up for that right now? Too lazy to open link - does his modeling factor in existing cases that haven’t been diagnosed - and their impending spread?

 

[BoilerBulldog]

 

[oldgoldblack]

Not sure If anyone has come across this guy’s model for predicting how COVID will spread and then slow down. Apparently it’s been tracking very closely so far. If his model is correct, US will peak on 3/28 and we will end up with around 1400 deaths on 180,000 infections. Anyone with a stronger stat background than me want to take a look and give me your thoughts? This is a far more positive outlook than many are predicting.

https://dispensationalpublishing.com/covid-19-scientific-prediction-model-march-23/

Haven’t opened the link but 1,400 deaths seems implausibly low—Italy had more than that just yesterday.

 

[mreman1220]

Fwiw Michigan hospitals believe the peak will hit around April 15. I think some death predictions are too high. This one is way too low.

 

[BoilerBulldog]

Fwiw Michigan hospitals believe the peak will hit around April 15. I think some death predictions are too high. This one is way too low.

Here’s his model for Michigan.

 

[kevindub]

Interesting that the peak in the US is before the rest in April since we are on the trail of the spread and one of the farthest behind the curve on supplies.

 

[AluminDallas]

Not sure If anyone has come across this guy’s model for predicting how COVID will spread and then slow down. Apparently it’s been tracking very closely so far. If his model is correct, US will peak on 3/28 and we will end up with around 1400 deaths on 180,000 infections. Anyone with a stronger stat background than me want to take a look and give me your thoughts? This is a far more positive outlook than many are predicting.

https://dispensationalpublishing.com/covid-19-scientific-prediction-model-march-23/

I like it, but there is one concern. His modeling is predicated on a bell curve distribution of daily cases. I read an article recently that suggested that there is no evidence to suggest this will have to follow a bell curve. I need to find the source because I think this is the great unknown from a data/modeling standpoint. Thanks for sharing.

 

[AluminDallas]

I like it, but there is one concern. His modeling is predicated on a bell curve distribution of daily cases. I read an article recently that suggested that there is no evidence to suggest this will have to follow a bell curve. I need to find the source because I think this is the great unknown from a data/modeling standpoint. Thanks for sharing.

Here’s some info on Farr’s law.

https://towardsdatascience.com/what...has-to-say-about-the-coronavirus-2473894f03c5

 

[Boilermouse]

Not sure If anyone has come across this guy’s model for predicting how COVID will spread and then slow down. Apparently it’s been tracking very closely so far. If his model is correct, US will peak on 3/28 and we will end up with around 1400 deaths on 180,000 infections. Anyone with a stronger stat background than me want to take a look and give me your thoughts? This is a far more positive outlook than many are predicting.

https://dispensationalpublishing.com/covid-19-scientific-prediction-model-march-23/

Wow is he optimistic! I have been looking at models that put the peak in late April or early May. I can’t believe our numbers will be that low either. Man, I hope he is right.

 

[JoeZap]

And @JoeZap.

There's not enough information on his initial assumptions to make an assessment on his models. However this statement is troubling: "When big enough, they average to a binomial distribution. (A “bell curve.”)" He's dealing with binomial distributions (diseased or not), (dead or not). With enough data the average of the binomial data asymptotically approaches a bell curve (normal or Gaussian distribution).

Because the Chinese hid the initial onsets of the disease and sent 100K (infected?) workers to Italy and Iran, their problems are exponentially worse.

I was at a testing site today. They tested 62 people. 58 had the flu and 4 had colds. It was fun. I was in the ER with chest pains and the coronavirus testing tents were right outside the window in my room. Turned out my electrolytes were out of balance.

 

[ahhculdee]

I really hope Trump sets a date for everything to open back up.

 

[BoilerBulldog]

There's not enough information on his initial assumptions to make an assessment on his models. However this statement is troubling: "When big enough, they average to a binomial distribution. (A “bell curve.”)" He's dealing with binomial distributions (diseased or not), (dead or not). With enough data the average of the binomial data asymptotically approaches a bell curve (normal or Gaussian distribution).

Because the Chinese hid the initial onsets of the disease and sent 100K (infected?) workers to Italy and Iran, their problems are exponentially worse.

I was at a testing site today. They tested 62 people. 58 had the flu and 4 had colds. It was fun. I was in the ER with chest pains and the coronavirus testing tents were right outside the window in my room. Turned out my electrolytes were out of balance.

Thanks. It will be interesting the next couple of days to see how the model trends.

 

[mreman1220]

Here’s his model for Michigan.

Like others said, I hope he's right but everything my fiance says about the situation at the local hospitals says otherwise.

 

[BoilerBulldog]

There's not enough information on his initial assumptions to make an assessment on his models. However this statement is troubling: "When big enough, they average to a binomial distribution. (A “bell curve.”)" He's dealing with binomial distributions (diseased or not), (dead or not). With enough data the average of the binomial data asymptotically approaches a bell curve (normal or Gaussian distribution).

Because the Chinese hid the initial onsets of the disease and sent 100K (infected?) workers to Italy and Iran, their problems are exponentially worse.

I was at a testing site today. They tested 62 people. 58 had the flu and 4 had colds. It was fun. I was in the ER with chest pains and the coronavirus testing tents were right outside the window in my room. Turned out my electrolytes were out of balance.

I'm a little confused, are you saying using a binomial distribution is incorrect? Per the poster above, it seems he's applying Farr's Law to this and the article AluminDallas posted highlights some of the risk associated with that. Alum's article's first coronovirus graph is almost identical to this guy's model.

 

[TODDBONZALEZ86]

I'm a little confused, are you saying using a binomial distribution is incorrect? Per the poster above, it seems he's applying Farr's Law to this and the article AluminDallas posted highlights some of the risk associated with that. Alum's article's first coronovirus graph is almost identical to this guy's model.

The statement JoeZap mentioned stuck out to me as well. The issue isn't in the application of 'Farr's Law' (which don't assume 'law' in this instance implies the usual meaning of law in science, because it's not) or if it's binomial data (it is), it's in his phrasing which could either be a simple typo/not proof reading issue or a lack of statistical knowledge. Either way, with zero information about how the models were put together, I wouldn't put too much stock into what they have to say at this point.
He also says he nailed the prediction about when the peak in Europe would be, which seems odd to me considering the date he mentions is over a week away and I haven't seen anyone say Europe has peaked.

 

[PurdueJ]

I really hope Trump sets a date for everything to open back up.

Haha how could he possibly do that with so much unknown???

 

[Purdue 95]

Haha how could he possibly do that with so much unknown???

I don't think anybody is going to draw a line in the sand and say, "OK, everybody quarantine through Sunday, then on Monday morning, it's back to business." They'll ease back into things, little by little. Otherwise, more and more businesses will begin to shut their doors and people will lose their jobs permanently.

Hopefully the drugs they are beginning to use will continue to show promise. Our ability to effectively treat the infected will make folks a helluva lot more comfortable.

 

[Boilermouse]

I don't think anybody is going to draw a line in the sand and say, "OK, everybody quarantine through Sunday, then on Monday morning, it's back to business." They'll ease back into things, little by little. Otherwise, more and more businesses will begin to shut their doors and people will lose their jobs permanently.

Hopefully the drugs they are beginning to use will continue to show promise. Our ability to effectively treat the infected will make folks a helluva lot more comfortable.

I think this is 100% correct. It is not binary. A real, effective treatment for this will be the cure for the fear AND the economy in my opinion. We aren’t there yet.

 

[stout1]

Haha how could he possibly do that with so much unknown???

Ignorance and lack of info has never stopped the man.

 

[laguna1]

 

[JoeZap]

I'm a little confused, are you saying using a binomial distribution is incorrect? Per the poster above, it seems he's applying Farr's Law to this and the article AluminDallas posted highlights some of the risk associated with that. Alum's article's first coronovirus graph is almost identical to this guy's model.

Saying that the binomial is a bell curve in incorrect. When you start plotting active cases (binomial occurrences) vs time you stop talking about statistical curves. These curves should be asymmetric. The curve starts at zero cases at zero time, rises to a peak and tapers off over time. The far side of the peak tapers off, sometimes to zero, but sometimes to a nonzero steady state.

 

[BoilerBulldog]

Saying that the binomial is a bell curve in incorrect. When you start plotting active cases (binomial occurrences) vs time you stop talking about statistical curves. These curves should be asymmetric. The curve starts at zero cases at zero time, rises to a peak and tapers off over time. The far side of the peak tapers off, sometimes to zero, but sometimes to a nonzero steady state.

Got it. Thanks.

 

[BoilerBulldog]

This guy updated his model today with new commentary for those interested. He did push the US peak date back one day to 3/29 due to the latest NY numbers.

https://dispensationalpublishing.com/covid-19-scientific-prediction-model-march-24/

 

[guidelinesa2]

This guy updated his model today with new commentary for those interested. He did push the US peak date back one day to 3/29 due to the latest NY numbers.

https://dispensationalpublishing.com/covid-19-scientific-prediction-model-march-24/

He thinks the peak new US cases per day will be on the 29th? Thats nuts.

It takes 14 days from lockdown for that to happen, based on Italy experience, and New York locked down 2 days ago. Then we are going to have to wait for it to roll thru the other big cities

 

[BoilerBulldog]

He thinks the peak new US cases per day will be on the 29th? Thats nuts.

It takes 14 days from lockdown for that to happen, based on Italy experience, and New York locked down 2 days ago. Then we are going to have to wait for it to roll thru the other big cities

Optimistic for sure. He stands behind how his models are tracking so far. I guess we will know in a couple of days. I find it interesting how closely the models are holding so far even with some smoothing done on his part.

 

[BoilerBulldog]

He thinks the peak new US cases per day will be on the 29th? Thats nuts.

It takes 14 days from lockdown for that to happen, based on Italy experience, and New York locked down 2 days ago. Then we are going to have to wait for it to roll thru the other big cities

I would caution comparing us to Italy. I don't think that's very realistic.

 

[CalBoiler]

I would caution comparing us to Italy. I don't think that's very realistic.

Italy = West Wuhan.

 

[greeneman]

Not sure If anyone has come across this guy’s model for predicting how COVID will spread and then slow down. Apparently it’s been tracking very closely so far. If his model is correct, US will peak on 3/28 and we will end up with around 1400 deaths on 180,000 infections. Anyone with a stronger stat background than me want to take a look and give me your thoughts? This is a far more positive outlook than many are predicting.

https://dispensationalpublishing.com/covid-19-scientific-prediction-model-march-23/

Given that he doesn't give his methodology there's no way to assess how good it is.

 

[BoilerBulldog]

Given that he doesn't give his methodology there's no way to assess how good it is.

He gives more detail in his blog today if you read through the comments for each graph.

 

[greeneman]

He gives more detail in his blog today if you read through the comments for each graph.

He still isn't saying how he did it, he just has little comments. His model may be awesome, or it may be trash, who knows cuz we can't see it.

 

[guidelinesa2]

I would caution comparing us to Italy. I don't think that's very realistic.

I don’t think we are that different

 

[ossim]

I don’t think we are that different

They have the second oldest overall population in the world next to Japan.

 

[guidelinesa2]

They have the second oldest overall population in the world next to Japan.

But there is more to it than age. There are plenty of comorbidities that the US had plenty of.

 

[dtimke]

If nothing else, it's nice to see someone is positive and I look forward to seeing if he comes out close. It's something to pass the time.

 

[oldgoldblack]

This guy updated his model today with new commentary for those interested. He did push the US peak date back one day to 3/29 due to the latest NY numbers.

https://dispensationalpublishing.com/covid-19-scientific-prediction-model-march-24/

Why are his initial projections 1,700 deaths in NY but only 1,400 in the U.S., am I misreading?

Either way his projections of around 1,400 total deaths in the U.S. will likely be the smoking gun that prove his models are off--probably by this weekend even.

 

[BoilerBulldog]

Why are his initial projections 1,700 deaths in NY but only 1,400 in the U.S., am I misreading?

Either way his projections of around 1,400 total deaths in the U.S. will likely be the smoking gun that prove his models are off--probably by this weekend even.

I’m pretty sure he said his total death number is getting adjusted tonight because of NYC. Which is also why he adjusted the peak day back one day. His model also shows around 13,500 new cases today. Which he expects to be slightly high.

Not by any means saying this guy is right. But I find it interesting to track from a statistical prediction accuracy standpoint.