Something fishy is going on with the IHME model

[UpTheMiddlex3Punt]

You know, the one here: https://covid19.healthdata.org/projections

They last updated on April first and said there would be an update on April 4. April 4 has come and (almost) gone and it still hasn't updated. They have removed the reference to April 4 as well.

I noticed there were some bad data in their set, especially when they were claiming Alabama would have something like 5,500 deaths by August, ahead of more populous and already harder hit states like Illinois.

Since this model is now being referenced all over the place as the authoritative model, are the researchers having an "Oh Crap" moment and realizing they better get their data right? Or have the more recent runs of the model widely diverged from what they had just a few days ago?

 

[SirBarksalot]

You know, the one here: https://covid19.healthdata.org/projections

They last updated on April first and said there would be an update on April 4. April 4 has come and (almost) gone and it still hasn't updated. They have removed the reference to April 4 as well.

I noticed there were some bad data in their set, especially when they were claiming Alabama would have something like 5,500 deaths by August, ahead of more populous and already harder hit states like Illinois.

Since this model is now being referenced all over the place as the authoritative model, are the researchers having an "Oh Crap" moment and realizing they better get their data right? Or have the more recent runs of the model widely diverged from what they had just a few days ago?

i was just wondering the same thing..and why they haven’t updated yet.
just looking at what 00 pointed out, with AL having 5k fatalities and NC with 1500 or so, that seems screwed up.

shttps://forums.sixpackspeak.com/showthread.php?211444-MS-COVID-Hospitalizations&p=1753802&viewfull=1#post1753802

 

[00Dawg]

Here's some more discussion on modeling this pandemic in general.

https://www.wired.com/story/the-mathematics-of-predicting-the-course-of-the-coronavirus/

The author is on the side of supporting the modelers, so there is a slant to it. Particularly with letting Imperial College off the hook, he buys right into their explanations without challenging them. It's obvious from their work and commentary after drastic changes that their model either much favors suppression over mitigation, or the virus simply isn't as impactive as they originally estimated.

I'm concerned the UW guys simply missed the mark as well. The author gives them a lot of leeway to miss, but Alabama still hasn't reached the low range of their "beds needed" number predicted for March 25th, and I don't see the model getting better. Supposedly the model was updated 4/1, so I can only surmise they're either just not paying attention or their model display can't be updated to show actual numbers. Depressingly, it's closest to accurate on deaths, but even that is currently off for Alabama by a factor of 5.

ETA: It may be that UW can't show actual results the display.
Here's their release page:
http://www.healthdata.org/covid/updates

I gotta admit, the 4/2 announcement looks like a lot of back-patting to me. There's no mentions of where they're missing the mark.
If you look at the notes below, it appears they date to after 4/2, having added Mississippi's stay-at-home order on 4/3. If Alabama's numbers drop dramatically in the next release, that will be a good indicator that they have that political decision as a huge factor in their formula. To me that would drastically decrease their credibility, as Alabama's stay-at-home order is not much of a change from the previous state of things.

 

[UpTheMiddlex3Punt]

A lot of the pressure to put the order into place came from people looking at this model.

Some of their other data looks to be wrong as well. They claim New York has only 718 ICU beds while governor Cuomo has stated on multiple occasions it's around 3000. Plus they state Alabama has 474, so they're telling us New York has less than twice as many ICU beds as Alabama.

 

[BhamDawg91]

On 4/1,

View attachment 15769
projected deaths were 7334. It’s now 5500. The peak was 4/22 now 4/19.

 

[bulldogbaja]

It now says updated on April 4th... but at the moment it sure looks like the same numbers it has been. It's definitely not updated for Mississippi right now.

 

[00Dawg]

Good news, everyone!

The Institute for Health Metrics and Evaluation vastly reworked their figures. You know, the one that countless government leaders have been relying on to make decisions and the national media has been shoving down our throats without a single objective discussion?
Turns out it was, way off for most states. Who knew? So now, instead of the low range of predictions being off for Alabama by a factor of 12 or more, it's only 65% higher than yesterday's actual figures.
We can at least rule out the "stay at home" trigger for the calculations, though, so I suppose it's encouraging they weren't trying to dictate policy. They just had a bad model.
The total predicted deaths for the US is now between 49,000 and 134,000.


Oh, and now only 7 states are predicted to run out of hospital beds, and 11 more out of ICU slots. Six of the 7 are in the Northeast, and the other is Nevada. Unless we're stupid enough to think we can totally stamp this out by everyone staying indoors, it's time to start planning for when certain demographics can start getting back to work in a lot of the country, and that needs to be in mid-to-late April for some places.

Now, we still have a crisis in the Northeast and you can't totally pull the lid off until we get some rigor around quick testing and effective treatments, but the good news is we have medical resources sitting on their hands across the nation right now that can be shifted to New York and surrounding environs.

 

[BhamDawg91]

Alabama now down to 923 projected deaths by Aug 4. That’s a significant difference from 7k. I would say the initial model was way off.

 

[EagleDawg97]

Wow, pretty unbelievable update for Alabama. It’s almost like the initial model was way off or something. No big deal, we have only destroyed a 22 Trillion dollar economy due to what this and other models said.

 

[40 acres and a mule]

The Institute for Health Metrics and Evaluation vastly reworked their figures. You know, the one that countless government leaders have been relying on to make decisions and the national media has been shoving down our throats without a single objective discussion?
Turns out it was, way off for most states. Who knew? So now, instead of the low range of predictions being off for Alabama by a factor of 12 or more, it's only 65% higher than yesterday's actual figures.
We can at least rule out the "stay at home" trigger for the calculations, though, so I suppose it's encouraging they weren't trying to dictate policy. They just had a bad model.
The total predicted deaths for the US is now between 49,000 and 134,000.


Oh, and now only 7 states are predicted to run out of hospital beds, and 11 more out of ICU slots. Six of the 7 are in the Northeast, and the other is Nevada. Unless we're stupid enough to think we can totally stamp this out by everyone staying indoors, it's time to start planning for when certain demographics can start getting back to work in a lot of the country, and that needs to be in mid-to-late April for some places.

Now, we still have a crisis in the Northeast and you can't totally pull the lid off until we get some rigor around quick testing and effective treatments, but the good news is we have medical resources sitting on their hands across the nation right now that can be shifted to New York and surrounding environs.

You know you can't state this, right? This is the apocalypse, dammit!*****

 

[EagleDawg97]

2 million dead and 18 months of shelter in place is the only acceptable response to this virus right? Anything other than that response, and you just don’t value human life. Did I get the proper way to respond to any encouraging data right?

 

[40 acres and a mule]

I think we agree with each other.

 

[engie]

That model has been a disaster throughout. At least they revised it from being off by a factor of 10 to off by a factor of two.

 

[msstate7]

Yeah, let's not celebrate just yet. It's pretty obvious that our experts are pretty terrible at projections

 

[jethreauxdawg]

Are we still supposed to yell at the people who have been “saying calm down it won’t be that bad”?

 

[WutheringDawg]

I can't remember what Mississippi's projection was, but it was at or near 1k, and is down to a projected total under 300 as of this morning's update.

 

[DawgInThe256]

We've been on a statewide lockdown since 5PM Saturday, so I guess it's working ***

 

[SirBarksalot]

Yeah, this thing has done a 180 for the most part in many places. Is it because the model is taking into account more stringent distancing measures?
It’s got Louisiana already past peak, so that should be one to watch in the next few days to see if it’s tracking accurately no?

 

[msstate7]

We've been on a statewide lockdown since 5PM Saturday, so I guess it's working ***

Damn right it is... Tate on schedule to save ~ 700 lives, and trump on schedule to save 1.8 to 1.9 million lives

 

[00Dawg]

Part of me wanted to think it was the measures...

and what they had built into their formulas. Unfortunately, you can look at when different states implemented the various options they included, and see that those weren't large factors.
They simply missed.

 

[LandsurveyorDawg]

When will deadheaddawg show up to tell us we are all idiots? People all over the country are calling the media on their lies. Videos of citizens showing up at emergency rooms and "Testing Centers" and them being empty after the "Media" ran reports of them being mobbed with 4 hour waits. All of the "Models" have been so badly exaggerated that one could almost call it a hoax (the buzz, not the virus--that is for certain real.) Reports of listing every death because of covid. I hear it is a miracle that deaths caused by the flu and pneumonia have dropped of the map in New York City. **

 

[40 acres and a mule]

and what they had built into their formulas. Unfortunately, you can look at when different states implemented the various options they included, and see that those weren't large factors.
They simply missed.

I read an article that said the model was designed with social distancing measures accounted for. I don't know for certain that is accurate.

 

[Shooter McGavin.sixpack]

When the imperial model first adjusted their findings, Dr Birx was optimistic. Then she sprung the 200K dead even if we do everything almost perfectly. Now this. All I wanted was for them to show their work, but they never would. It was always just take our word for it. Couple that with their dragging their feet on getting a representative data sample and antibody tests. It's a mess.

 

[fishwater99]

That model has been a disaster throughout. At least they revised it from being off by a factor of 10 to off by a factor of two.

Crazy that all of our policy is off one model.
Have you seen how many models that are out there for Hurricanes and how much they differ.
What if we only went off one model for Hurricanes?

 

[Leeshouldveflanked]

I would like to see a graph on “all deaths”... I would suspect that less people have died during this time due to staying safe than average due to accidents, wrecks etc

 

[LandsurveyorDawg]

"I would like to see a graph on “all deaths”... I would suspect that less people have died during this time due to staying safe than average due to accidents, wrecks etc"

No doubt. I would love to see that as well.

 

[HotMop]

View attachment 15769
projected deaths were 7334. It’s now 5500. The peak was 4/22 now 4/19.

April 5th update...Now it says peak is 4/15?

 

[Cooterpoot]

I see this model has been revised to estimate 60,000 projected deaths in the U.S. now. That's a far cry from the "millions" and "200,000 if we do things perfect".

 

[bulldogbaja]

I assume you are posting this is a celebration- in which case I agree!

 

[turkish]

I see this model has been revised to estimate 60,000 projected deaths in the U.S. now. That's a far cry from the "millions" and "200,000 if we do things perfect".

Hold your horses for now. It predicts LA has peaked but that’s not yet a proven fact. I certainly hope you’re right on, though.

 

[Cousin Jeffrey]

Is it because the model is taking into account more stringent distancing measures?

The IHME model has always assumed full social distancing measures were in place.

So it has just been wrong. And we’re doing our best to destroy an economy based on it.

 

[WeWonItAll(Most)]

Wow, pretty unbelievable update for Alabama. It’s almost like the initial model was way off or something. No big deal, we have only destroyed a 22 Trillion dollar economy due to what this and other models said.

The hardline social distancers and those predicting doom will soon turn and begin patting themselves on the back saying that the reason it wasn't that bad was because of the social distancing measures we took and their ability to "scare people into taking this seriously"

Computer modeling with very limited data is infallible and unquestionably reliable****

 

[DoggieDaddy13]

Okay, so we've missed out on a lot of sports, but the ECONOMY has not been destroyed.

You guys are funnier than the "everybody's gonna die" crowd. Way funnier actually. The "everybody's gonna die" crowd's kinda morbid. More of that dark humor, I reckon.

 

[hatfieldms]

It maybe hasn’t been destroyed completely yet but you are fooling yourself if you do not think it is extremely damaged. Hopefully it will be able to bounce back fairly quick

 

[Shooter McGavin.sixpack]

Limited data has always been the problem. In the beginning, they had no choice but to employ all these lockdowns. There was so much unknown about the virus. What we now know is still not where it should be. This is where I bang the we need antibody tests and a representative data sample drum..again, but we are gathering more info every day. We can't undo the damage done, but hopefully we see more targeted measures moving forward as appropriate.

 

[Cooterpoot]

Okay, so we've missed out on a lot of sports, but the ECONOMY has not been destroyed.

You guys are funnier than the "everybody's gonna die" crowd. Way funnier actually. The "everybody's gonna die" crowd's kinda morbid. More of that dark humor, I reckon.

Wait til you see how many businesses don't come back and how hard the pending recession hits people. The economy is barely limping along right now and losing more small business is terrible for everyone.

 

[DoggieDaddy13]

"We will OPEN UP OUR GREAT COUNTRY.... and our economy will BOOM, perhaps like never before." The real Donald J. Trump.

He's got this.

 

[WeWonItAll(Most)]

Limited data has always been the problem. In the beginning, they had no choice but to employ all these lockdowns. There was so much unknown about the virus. What we now know is still not where it should be. This is where I bang the we need antibody tests and a representative data sample drum..again, but we are gathering more info every day. We can't undo the damage done, but hopefully we see more targeted measures moving forward as appropriate.

I just don't see any indication of this happening. I feel like I see more talk about still being in total lockdown 6 months from now than I do talk about efforts to get the world moving again. No talk about when the lockdown will be lifted or what the goal is for ending it. No talk about a more targeted approach. Just talks about what draconian measure should be taken next as we continue to race to the bottom and of more gargantuan emergency government spending.

Maybe I'm missing it

 

[EagleDawg97]

New leading model update says 60,000 Covid deaths in the US.

Interesting factoid according to the CDC the total influenza deaths for 2017-2018 season in the US was 61,000.

 

[Cooterpoot]

I can't believe this is similar to the flu***