Yeah, U.S. will hit 1,400 deaths by this weekend, if not before...not that I want to closely watch that number.
COVID-19 Statistical Prediction Model
- Thread starter BoilerBulldog
- Start date
I’m “enjoying” this site. Looks like he’s going to be close on new cases for today unless the numbers skyrocket late. The key will be when he hits the top of of his parabolas around the 29th. Will the numbers continue to go up? Hoping beyond hope they follow his predictions. I also like that he doesn’t seem to be full of himself and will readily admit if he’s wrong. Talks pretty plain too.
His new post is up. He's pushed the peak back one more day to 3/30 and is having to rework his death projections due to NYC. But overall, he dropped our overall infection number by 20K and thinks we are still tracking toward a binomial distribution.
Dr. Frank answers the question, “Why will the cases peak rather than continue exponentially?”
Dr. Frank answers the question, “Why will the cases peak rather than continue exponentially?”
Of course. Shouldn't you update the algorithm though? I'm not sure why you wouldn't replace projected data with actual results and then evaluate the performance of the algorithm.
As long as you do that fine. It didn't appear that was what he was doing.
Nobody is going to nail this thing perfectly. But it won't be too hard to go back and see how this prediction model stacks up against the doomsday prediction models.
It would be interesting to see how many adjustments he has to make based on his algorithm projecting too low. I also don't see how this will level back down to zero until there's an actual cure/vaccine. If CA blows through 3/26 and still has an upward trajectory, will he acknowledge that?
Please stop using Italy as a comparison to the US. There is evidence that they are overstating the number of deaths by attributing the virus as the cause of death for anyone dying in a hospital housing COVID-19 patients.
This is an article form yesterday's Basel paper. Open the link in Chrome and translate it.
https://www.bzbasel.ch/schweiz/hohe-lebenserwartung-viele-tote-wie-steht-es-um-die-schweiz-137335320
A paragraph from the article. It refers to the death count of one day in Italy:
"However, the statistical numbers have to be put into perspective, fortunately being old alone is not a corona death sentence. Because it can be heard from the hospitals and from infectiologists that it primarily affects people who are at the end of their lives anyway. So patients with previous illnesses who would otherwise have died within a short time. To date (should be "this date"), only three deaths have been identified in Italy that died simply because of the novel virus. Every death is a tragic individual fate, but it is not decisive whether the corona virus or another pathogen is responsible for it two weeks later. This only affects the corona virus death statistics. In addition, age is not the only decisive factor for the corona risk, but health and lifestyle of the elderly."
Switzerland, Germany and Austria are much better comparisons. Focus on the last two columns. Infection rate and death rate. Switzerland's infection rate is very similar to Italy's while the death rate is much lower. CH updates its numbers throughout the day, IT updates in the evening.
This is an article form yesterday's Basel paper. Open the link in Chrome and translate it.
https://www.bzbasel.ch/schweiz/hohe-lebenserwartung-viele-tote-wie-steht-es-um-die-schweiz-137335320
A paragraph from the article. It refers to the death count of one day in Italy:
"However, the statistical numbers have to be put into perspective, fortunately being old alone is not a corona death sentence. Because it can be heard from the hospitals and from infectiologists that it primarily affects people who are at the end of their lives anyway. So patients with previous illnesses who would otherwise have died within a short time. To date (should be "this date"), only three deaths have been identified in Italy that died simply because of the novel virus. Every death is a tragic individual fate, but it is not decisive whether the corona virus or another pathogen is responsible for it two weeks later. This only affects the corona virus death statistics. In addition, age is not the only decisive factor for the corona risk, but health and lifestyle of the elderly."
Switzerland, Germany and Austria are much better comparisons. Focus on the last two columns. Infection rate and death rate. Switzerland's infection rate is very similar to Italy's while the death rate is much lower. CH updates its numbers throughout the day, IT updates in the evening.
Agreed. I just appreciate a math/stat approach to the hysteria. He's readily acknowledged errors he's had to correct, primarily due to NYC. And he's pushed back the projected peak day by one day the last two days. I think he's way optimistic but too many people fail to recognize there's a way to predict infection rate that's not perfect but should take some of the hysteria away. Virus spread starts exponentially but eventually pivots to a normal curve. He's just trying to predict the inflection point. And he's very open about what he's doing and how he's doing it. I find that interesting from a purely mathematical standpoint. And if I'm honest, it's given me a little bit of hope the past couple of days with the overwhelming bad news. I'm not banking on this thing peaking within the next week. But I find myself excited to see how the model is adapting each day.
I like this "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)." It is my understanding that all averages follow a bell shape curve (with enough data), but perhaps there exists some that doesn't? I almost posted some sample sizes necessary to provide a confidence level of catching x% of such and such with the binomial and Poisson. I'm not second nature on this anymore, but really enjoyed reading your post and using words like binomial...Gaussian and so forth!
Right now we have a lot of subsets that contain different sources of variation confounded in the "grouping or categories" of the subsets and cannot break down further (like smoking etc...other poorer health conditions for weighting if that is ever possible). Seems like many years ago we would take the np and if over 10, some wanted 20 to get enough samples for each category (in testing for significance) and take the arcsin of the percent defective as a means of normalizing the data and then apply normal Analysis of Variance for significance. Either way, I feel younger just reading your post...but not young enough I want to relearn all I forgot!
He's posting on Facebook. The website (posted at the start of this thread) that is aggregating all his FB content just hasn't updated.
At least I think this is what you're seeing. Best place is to look Douglas G Frank up on FB.
Michigan was a surprise jump in cases today and there was some dialogue on it. The rest of the states were tracking as expected yet.
Nice catch! I didn’t think any numbers from yesterday would have thrown him off enough to stop. Glad to see he’s still plugging away.
A model cannot be updated daily, thats not a model, thats a graphing exercise.
Amateurs cannot model this thing with our data, without knowing the exact extent of testing, as well as the unknown percentage of patients without symptoms.
Unless this guy is a genius, he’s wasting his time
Amateurs cannot model this thing with our data, without knowing the exact extent of testing, as well as the unknown percentage of patients without symptoms.
Unless this guy is a genius, he’s wasting his time
If the doomsday numbers don't happen, you'll never hear any pushback from the MSM.
If the doomsday numbers don’t happen doesn’t that mean that something went well? Why would anyone make a big deal that the doomsday numbers were wrong in that case?
He's not adjusting the models daily...he's charting actuals on to of the models already created and he is actually deliberate about only changing when there is sufficient data making it clear the model needs to be fixed and then it's a discussion of why. No ego.
Models are guesses at this point, and pretty meaningless until there's a vaccine.
Much about the coming years, including the frequency, duration, and timing of social upheavals, depends on two properties of the virus, both of which are currently unknown. First: seasonality. Coronaviruses tend to be winter infections that wane or disappear in the summer. That may also be true for SARS-CoV-2, but seasonal variations might not sufficiently slow the virus when it has so many immunologically naive hosts to infect. “Much of the world is waiting anxiously to see what—if anything—the summer does to transmission in the Northern Hemisphere,” says Maia Majumder of Harvard Medical School and Boston Children’s Hospital.
Second: duration of immunity. When people are infected by the milder human coronaviruses that cause cold-like symptoms, they remain immune for less than a year. By contrast, the few who were infected by the original SARS virus, which was far more severe, stayed immune for much longer. Assuming that SARS-CoV-2 lies somewhere in the middle, people who recover from their encounters might be protected for a couple of years. To confirm that, scientists will need to develop accurate serological tests, which look for the antibodies that confer immunity. They’ll also need to confirm that such antibodies actually stop people from catching or spreading the virus. If so, immune citizens can return to work, care for the vulnerable, and anchor the economy during bouts of social distancing.
Much about the coming years, including the frequency, duration, and timing of social upheavals, depends on two properties of the virus, both of which are currently unknown. First: seasonality. Coronaviruses tend to be winter infections that wane or disappear in the summer. That may also be true for SARS-CoV-2, but seasonal variations might not sufficiently slow the virus when it has so many immunologically naive hosts to infect. “Much of the world is waiting anxiously to see what—if anything—the summer does to transmission in the Northern Hemisphere,” says Maia Majumder of Harvard Medical School and Boston Children’s Hospital.
Second: duration of immunity. When people are infected by the milder human coronaviruses that cause cold-like symptoms, they remain immune for less than a year. By contrast, the few who were infected by the original SARS virus, which was far more severe, stayed immune for much longer. Assuming that SARS-CoV-2 lies somewhere in the middle, people who recover from their encounters might be protected for a couple of years. To confirm that, scientists will need to develop accurate serological tests, which look for the antibodies that confer immunity. They’ll also need to confirm that such antibodies actually stop people from catching or spreading the virus. If so, immune citizens can return to work, care for the vulnerable, and anchor the economy during bouts of social distancing.
The interesting MSM dynamic I see is they are constantly saying Trump should listen to his scientists (Fauci and Birks).
Yet they largely ignore Fauci and Birks analysis of data in their own reporting and instead rely on their own medical analysts and Cuomo.
There is an incredible disparity between regions - countries in Europe and states in the US that doesn't allow cross comparison of data to form a universal model/prediction.
NY skews the US data. With New Jersey, the 2 states are more than half of all reported cases. The infection rate in NYC looks like nothing else, stateside. What would one expect when you stack >8 million people in a relative cracker box with a population that relies on public/common transportation, massive amounts of shared common space, and huge influx of non-NYCers everyday. That probably pushes the daily, real population solidly over 9 million, prior to 10 days ago. A city block there has more people than most towns in the midwest.
Northwest Indiana has ten or so hospital facilities across a few counties. Those three counties have 24 positive results (3/25). 2 of those counties have had drive-thru testing since last week. There are reportedly, less than a third of the 24 under hospital observation. Others are self quarantined. Now, in reality, there are probably more than 24 people carrying, but there are a whole lot of people I know self-quarantining at this point. Around half the counties in Indiana have less than 2 positive cases.
Illinois has nearly 2,000 cases, 19 deaths, the vast majority of cases/deaths are in the Chicago metro area (6+mil), which is substantially geographically, larger than NYC. The infection rate and pure number are no where near NYC. Nearly every county outside of the Chicago collar have less than 5 cases, with the vast majority reporting 0.
There is no universal model that applies across all states, let alone countries. The curve flattening to preserve hospital space is a much bigger issue in densely populated areas.
NY skews the US data. With New Jersey, the 2 states are more than half of all reported cases. The infection rate in NYC looks like nothing else, stateside. What would one expect when you stack >8 million people in a relative cracker box with a population that relies on public/common transportation, massive amounts of shared common space, and huge influx of non-NYCers everyday. That probably pushes the daily, real population solidly over 9 million, prior to 10 days ago. A city block there has more people than most towns in the midwest.
Northwest Indiana has ten or so hospital facilities across a few counties. Those three counties have 24 positive results (3/25). 2 of those counties have had drive-thru testing since last week. There are reportedly, less than a third of the 24 under hospital observation. Others are self quarantined. Now, in reality, there are probably more than 24 people carrying, but there are a whole lot of people I know self-quarantining at this point. Around half the counties in Indiana have less than 2 positive cases.
Illinois has nearly 2,000 cases, 19 deaths, the vast majority of cases/deaths are in the Chicago metro area (6+mil), which is substantially geographically, larger than NYC. The infection rate and pure number are no where near NYC. Nearly every county outside of the Chicago collar have less than 5 cases, with the vast majority reporting 0.
There is no universal model that applies across all states, let alone countries. The curve flattening to preserve hospital space is a much bigger issue in densely populated areas.
agree with both of you,...additionally from what I am hearing and seeing in reports from medical professionals, I'd say their are at least two very likely effective treatments to stop and reverse the symptoms I.e. disease progression. Medical community are using both and seeing very good results within days of administration. Medical community won't wait on "official " FDA approval...they will adopt it ASAP if they see improved clinical results.
Considering the H1N1 virus killed around 12,000 in the US, I would agree the number will rise. Hopefully not to the H1N1 numbers.
The Swiss government just committed CHF 100,000,000 to develop a COVID-19 vaccine as well as manufacturing and distribution infrastructure for use only inside Switzerland. They have learned their lesson very quickly as DE is withholding masks and other protective items they normally send here.
I posted earlier. There is a reason the Illinois numbers are a trickle - constantly 13% positive. This already ran through Chicago metropolitan area in January. There are 1000s of Chinese students who went home for break and came back in early January. There are entire condo developments in south loop that are Chinese owned. China Town has really expanded and is vibrant. And Chinese nationalists have moved into burbs to divest and get into great school districts. This has been happening since 2014. Lots of real estate articles about the influx of Chinese buyers. Thankfully for many markets. So anyone who thinks the first Chicago case was late January is simply wrong. It’s just common sense.
ONE HUNDRED MILLLLLLLLLLLLION FRANCS!
I'm glad people are doing something but its not really all that news worthy
I REALLY don't want to argue about this, but if you have the former head of Obama's ObamaCare push publicly stating on Twitter that he predicts 2,000,000 Americans will die from this, and that number (hopefully) is way off, don't you think statements like that are irresponsible and fear-mongering? Why is it perfectly ok to blast Trump for trying to be ++++ about it but then the media gives guys like this a pass. I think statements like that are far more damaging to the public. Does that make any sense?
Yet you replied.
Switzerland's population is about 8.5 million. Has the US committed $4,000,000,000 for developing a COVID-19 vaccine and the related manufacturing and distribution infrastructure? Nope, too busy giving money to the Kennedy Center for the Arts and other BS having nothing to do with helping.
BTW I just had a close MD friend of mine tell me that I should NOT be at work so this is quite the bind.
