COVID-19 Statistical Prediction Model
- Thread starter BoilerBulldog
- Start date
Yes, I know that part. I’m talking about overrun hospitals, total number of cases and deaths, etc.
Is this a change in their outlook? I feel like lots of people I interact with (county health department, IN DOE, governor’s office, ISDH, our local hospital) have been concerned for a couple of weeks.
This website appears to be much more accurate which has the peak still 17 days away.
https://covid19.healthdata.org/proj...2CfOamy_FUPlUND-FTbouW6PrGLtAVPOb96t6ZpO6FGG4
https://covid19.healthdata.org/proj...2CfOamy_FUPlUND-FTbouW6PrGLtAVPOb96t6ZpO6FGG4
He's updated his death numbers to reflect NYC:
And here's his explanation:
Covid-19 "USA Deaths Tracking with NY"
This graph is really revealing. And more importantly, it confirms an hypothesis we had a couple days ago. Cha-ching!
I have changed nothing about any of the models. Think of this as merely "accounting." This is the first time I have ever looked at this plot, because I just made it. I've been stewing over the right way to do this for a couple days, and it was the first thought on my mind when I woke up. (The power of the subconscious mind...)
Let me walk you through it. I usually don't like putting so many traces on a single graph for public viewing. It scares people off. But it is worth it, trust me. Take your time, digest each plot. Look at the scale for it, and don't move on until you get it intuitively.
Let's start with the axes. The left vertical axis is for all "Deaths/day" curves, the right vertical axis is for the two "sigmoids," or total death curves that finish into it. (Near 1400 and 2900.)
Let's start with the red dashed plot with the red dots, USA Reported deaths per day. The scale for this is in red, on the left. It is real data, so it is noisy. But it is reality. (Always start with the DATA!!)
The pink shaded line that is tracking the red-dots-curve is the sum of my original projection for the whole USA (deaths/day) plus the model predictions for NY (deaths/day). This "sum of models" is closely tracking what we are **actually** observing. (It is the sum of the light gray and blue peaks.)
Next, the solid dark gray line is my original model death tally estimate for the whole country (finishing on the right, near 1400 total death tally). The light gray peak is the cases/day for that model (use left scale).
The light blue curve is the peak corresponding to the deaths/day predicted by our NY model.
When you add the original model to the NY model, you get the dark red plot, finishing just under 3,000 total deaths.
The dark red dots and dashed curve is the total reported death tally for the entire US (actual data). Note the "inflection point" in the curve corresponding to when the NY death peak starts growing and the trace begins diverging from my earlier estimate. (This is exactly what a "secondary infection looks like in other countries, eg Italy.)
I am not saying NY is a secondary infection. I'm saying, when you consider the data and models separately *and* together, this approach provides valuable insight into what is going on in our country.
I will update this graph every day for a while. I might simplify it too... but there is so much good information in here, I don't want to leave anything out. I will think on it.
What I love about this, is that we hypothesized a couple of days ago that this way of thinking about the country would match the data... AND IT DOES. An a priori hypothesis confirmed, making us more confident in our models. This is the scientific method.
It also reveals that my original assumptions (and Midwestern bias?) about our country were likely incorrect, and that I will need to revise my assumptions. Note that I am being careful to let the data guide my thinking. Make an hypothesis, then test it. Learn. The scientific method works.
So looks like he just doubled his original estimates of total deaths. Common sense said from day one his death predictions were completely implausible. Even after this adjustment his projections are still implausible, but I’m sure when the real numbers start getting close to his new revised number, he will adjust his graph again.
@Tony79 this isn’t an argumentative post, it is a genuine question. It is interesting to me to watch people’s views evolve as we have gotten in to the midst of this.
Total Cases and Total Deaths seems to be tracking as predicted. The cases/day had one day we’ll outside the curve. We have different definitions of “blown up”.Dude's California model is 100% BLOWN UP
Actually Cal, the last 4 days have all been higher than his model predicted, and the overall case volume is now higher as well it's just harder to see because of his scale. And if you could actually zoom in on the total deaths it's higher than his model too, it's just hard to see because he has too many things on one graph with different axes. Also, the deaths will likely see massive growth because they will usually lag reported cases. Reported cased per day doubled in 1 day, the deaths will probably follow.
I've been following this guy for a while, I do this type of stuff for a living though not with health data. He's a smart guy obviously, more prominent in the field than me by far, but he has fallen into a common trap. GIGO. With the scarcity of good testing, the data is dirty and incomplete, and he tried to build a model on awful data.
Over the last couple of weeks or so, it looks to be about half of the time his total case numbers were lower than expected. Also, to me, prior to yesterday’s outlier, the cases/day seemed to be leveling off, as the model predicted, albeit a bit higher than predicted.
Again, is the real data smack dab on the model? No. Is is pretty close? I’d say it is, so far. Let’s see what happens going forward.
What model or website do you recommend following? If the answer is none of them because nobody has access to good enough data then I look at this guys approach as the way I'd like to follow and learn about what's going on. I also like that he breaks out by states, etc as it is easier to see how different parts of the country are doing and highlights the fact that the US is really more like a conglomerate of countries than a single case. Too many comparisons of the US to smaller countries that are great for headlines but not much else. Why can't we compare the US to all of continental Europe?
He's not promoting his model as being locked in on x date and will hold true until the end while everyone else is wrong. He has a starting place and is tracking to see how it does. When the data says to revise, he does. It's not about being right but rather sticking to things that are more than opinions and conjecture and he does some teaching along the way where I feel like I'm gaining something.
All that said - if you've got someone else that is nailing this whole thing I'm all in!
"He's not promoting his model as being locked in on x date and will hold true until the end while everyone else is wrong. He has a starting place and is tracking to see how it does. When the data says to revise, he does. It's not about being right but rather sticking to things that are more than opinions and conjecture and he does some teaching along the way where I feel like I'm gaining something."
Agree 100% with this. Am enjoying his analysis. He's definitely not a "I know what I'm doing and you don't guy." If he's off he'll say so and try to learn from it to make his model better. He constantly talks about the scientific method. None of the state models have really reached their zenith yet (Minnesota maybe) and won't until mid-next week. Too early to tell. He definitely has the teaching background and often relates past teaching moments with his students.
Agree 100% with this. Am enjoying his analysis. He's definitely not a "I know what I'm doing and you don't guy." If he's off he'll say so and try to learn from it to make his model better. He constantly talks about the scientific method. None of the state models have really reached their zenith yet (Minnesota maybe) and won't until mid-next week. Too early to tell. He definitely has the teaching background and often relates past teaching moments with his students.
His california model had today as the first day on the decrease. Instead case volume per day more than doubled in one day.
I have not talked to these folks for while so I don't really know. I think given there is such a broad array of symptoms on top of the crazy 1% of the patients who just crump literally within a couple of hours. Those things really bother them.
I did not take it that way at all. My overall take is this mass criticism of Trump is not helping and trying to order up all these supplies takes time. I guarantee you a) if the gov't had spent all this money up front and/or b) if the hospitals had magically had the foresight to spend all this $$$$$, there would have been plenty of people that would have criticized all the spending. I don't think anybody handled this well at the start including NYC politicians. I think, in general, the pressers are a great way to be transparent. I just think Trump should let the experts handle them.
No business/government can mitigate every risk.
Here is what he said about California. Take it for what you will.
Just added data.
Huge jump in detected cases. When we see a huge jump in cases, but no change in deaths, that is good. It usually means a testing anomaly.
The model predicts 82 deaths *tomorrow* and that is what the deaths were *today*. An increase of 17 more deaths today is serious, but not reason for panic. In California, ten people die each day in car accidents.
So for now, I'd stick with testing anomaly; not an inflection point. But I am going to be paying attention tomorrow. My mom lives there, and she is vulnerable.
If the governor is trying to scare us; it is working. The initial projection of 115 is probably going to be low, because we are just only over the projected peak. Probably double. On a log scale, that is close. Let's pray the count stays low. Prayin' for you, Mom.
(You can see my personal biases at work here... gotta be extra careful. Go with the math; go with the math; go with the math...)
As you said, GIGO. Anyone thinking this guy's models were going to be accurate at this stage was engaging in wishful thinking and nothing more.
Last edited:
That is the scary thing. Given the symptom anomalies one would think we are seeing some mutations. Not everyone is getting Covid-19. Some people are getting Covid-19.1, 19.2, 19.3, etc.
You may recall that for about 4 weeks, after I had my failed dental implant removed, for the first time since I was a HS soph (1973) I felt horrible - dull headache, did not have my legs under me, and just general weakness - could not have gone to the gym if my life depended on it. Then at the end I had what I thought was rotovirus - terrible nausea, vomiting, blinding headache for 2 days, 103 fever, etc. Finally starting about two weeks ago I came around.
My MD friend I talked to this AM said I very could have had Covid then. I might end up getting titres drawn here in a few weeks and see if I have antibodies to it. Not everyone who gets it has respiratory symptoms.
Of course all viruses mutate constantly, but I’ve seen several reports suggesting this virus is fairly stable as these things go. A Google search on Corona virus mutation will bring up links for Fox News or the Washington Post for whatever ones personal taste is.
This was the closest thing I could find to the original source
https://www.prnewswire.com/news-rel...s-to-coronaviruss-epidemiology-301029331.html
This would be good news for vaccine speed and effectiveness
Here we go folks !
100,000 cases already / 1500 dead.
Think April looks like: 500,000 cases / 10,000-15,000 dead
Hoping . . it peaks in May
Michigan and others already drafting policies of caring for those "most likely" to survive.
PS If we go back to work by Easter, all bets are off. We won't even see the spring break impact by then
100,000 cases already / 1500 dead.
Think April looks like: 500,000 cases / 10,000-15,000 dead
Hoping . . it peaks in May
Michigan and others already drafting policies of caring for those "most likely" to survive.
PS If we go back to work by Easter, all bets are off. We won't even see the spring break impact by then
About a month ago, I thought March would see 50,000 cases and 1,000 dead. I underestimated how fast our medical supplies would run out and how fast our health care workers would be infected.
If we don't see a slowing or leveling off of new cases between now and then, we'll hit 500,000 by next Friday. That's April 3rd.
That should give Jim time to identify all the high-risk folks and dig a moat around them.
That is a higher rate of infection than what has been found in USA. Pretty sure USA is still well below 10% in terms of people who were tested actually have it.
The virus mutates very slowly which is why some are hopeful for a vaccine quicker. There are 8 strains circulating now. If interested here is a USA Today article
https://www.yahoo.com/news/8-strains-coronavirus-circling-globe-182338255.html
The situation in Michigan keeps sounding more and more like the apocalypse. They are bringing in the army to build a field hospital.
Last edited:
I know this is probably an inappropriate joke given the situation, but i did think it is funny ...
Last edited:
Unfortunately, I think the apocalypse could be Florida - Miami is now becoming a new hot spot and the demographics are horrible for this virus
Anybody that nails it exactly is going to be somewhat lucky, as the inputs required a lot assumptions. An article I read the other day looked at 18 infections disease expert models the average predicted death was around 200K, with the most extreme 1.2M.
It will be interesting to see how this guys model stacks up against the experts predictions. Would think there is a decent chance he ends up being more accurate than the experts.
I do wonder why this is not being done on a more proactive basis.
The army along with reserve and NG components can build many of these units.
Get proactive and build them in major city/metro areas.
No need to wait, it is when, not if.
Someone said a bunch of New Yorkers fled to Florida to their winter homes to avoid the spread there. Just took it to a new location
Well sure, and each area is going to react differently. I think anyone with some sense would know that if it got to NYC and had a chance to spread before meaningful actions were taken that - duh - NYC would be the worst place in the US. That doesn't mean someplace else won't surpass it and throw models today all to hell.
I believe CDC's worst case was something like 120,000,000 cases with 1% morbidity, so that 1.2M number would make sense. Trump's doctor got up there yesterday and - according to some - "discredited" those numbers... but I guess we'll see. It's alarming to me how few people are still actually taking this seriously.
Just from observations it appears to me there is a big disconnect between Fauci/Birx and Cuomo & his medical experts/analytics team
Ive seen this too. I read that the vaccine may be good for a few years, also saw good news that immunity is gained by survivors.
This is being done in NY snd CT, New York has converted the Gavits Center into a massive hospital, for starters.
More! NOW!
Too much of the country is STILL in denial about this. That’s how we got here.
Bad day in the US. It looks like right now our daily testing capacity is about 110k tests per day. I’m guessing its limited by the labs running the tests and not the kits themselves.
