Taking current state and applying future projections.

I don't understand how learned people can take them seriously. Have you ever moved cities? When you go to the grocery store the first time and don't know where shlt is, it takes for effing ever. 38 weeks later you can visualize the same tortilla shelf with your eyes closed.

These are smart people building models. And they flipping suck at it.

Models don't take into account:

1. Self social distancing in fear
2. Medical advancement
3. Therapeutic remedies being discovered and tested
4. Public adoption of PPE
5. Other changes in population behavior.

These are obvious things to even the most average layperson. But every model forecasts a sheep like populace acting like the cast of idiocracy. Do the scientists think the world populace has exactly zero survival instincts? Because that's how they model. It's freaking asinine.

It's also why I think this fear of a second peak or second wave in Winter 2020 is dumb. They act like we aren't going to learn anything in 8 months. Who the eff thinks that? Think of what we have learned in just the last six weeks! Are we going to stop learning for the next 30 weeks?

Have faith in the American system. Really smart people have the opportunity to become really rich. They will figure it out sooner rather than later.

"As soon as you attempt to measure human behaior, you change it making your measurements less valid"

Bonfire1996 said:

---

Taking current state and applying future projections.

I don't understand how learned people can take them seriously. Have you ever moved cities? When you go to the grocery store the first time and don't know where shlt is, it takes for effing ever. 38 weeks later you can visualize the same tortilla shelf with your eyes closed.

These are smart people building models. And they flipping suck at it.

Models don't take into account:

1. Self social distancing in fear
2. Medical advancement
3. Therapeutic remedies being discovered and tested
4. Public adoption of PPE
5. Other changes in population behavior.

These are obvious things to even the most average layperson. But every model forecasts a sheep like populace acting like the cast of idiocracy. Do the scientists think the world populace has exactly zero survival instincts? Because that's how they model. It's freaking asinine.

It's also why I think this fear of a second peak or second wave in Winter 2020 is dumb. They act like we aren't going to learn anything in 8 months. Who the eff thinks that? Think of what we have learned in just the last six weeks! Are we going to stop learning for the next 30 weeks?

Have faith in the American system. Really smart people have the opportunity to become really rich. They will figure it out sooner rather than later.

---

Yep. Variables always change. People adapt. Especially Americans. And thank God for that.

First of all, I would say that the people making these models do take all of those things into account. However, they have to make estimates at some of the values and some of the estimates may be off. That's why it's important for he model to be continuously updated as additional information comes to light and estimates can be better dialed in.

Second of all, I think the models have performed pretty well given the massive uncertainties we have been dealing with in terms of how many people really have this. Look at this model that was presented back in March at a WH Press Conference that Dr. Birx is talking to. Our daily death count for the country has followed this model almost perfectly.



Third of all, I would say that these models are all being designed for different purposes and it's important to understand the purpose of those models when you're reading into the data. The IHME model was designed to predict the maximum health care resources needed in order to try to forecast if we would run into shortage situations, from what I understand. If that's the case, the exact shape of the model or what it predicts in terms of total people dying is actually inconsequential.

Finally, even though models may be rife with uncertainties for a number of good reasons, they're still better than nothing. They not only allow us to make our best forecasts with the information we have now based on past events. It also allows us to do sensitivity testing on different policies and ideas.

FiveThirtyEight, which is a data journalism site headed by Nate Silver who is famous for his models and forecasting on subjects from sports to politics had a great article about why modeling COVID was so difficult.
https://fivethirtyeight.com/features/why-its-so-freaking-hard-to-make-a-good-covid-19-model/ I like their wrap up statement in this case:

Quote:

---

Think of it like making a pie. If you have a normal recipe, you can do it pretty easily and expect a predictable result that makes sense. But if the recipe contains instructions like "add three to 15 chopped apples, or steaks, or brussels sprouts, depending on what you have on hand" well, that's going to affect how tasty this pie is, isn't it? You can make assumptions about the correct ingredients and their quantity. But those are assumptions not absolute facts. And if you make too many assumptions in your pie-baking process, you might very well end up with something entirely different than what you were meant to be making. And you wouldn't necessarily know you got it wrong.

Over the next few months, you are going to see many different predictions about COVID-19 outcomes. They won't all agree. But just because they're based on assumptions doesn't mean they're worthless.

"All models are wrong, it's striving to make them less wrong and useful in the moment," Weir said.

We're hungry, so somebody has to do some baking. But be sure to ask what ingredients went into that pie and in what quantities.

---

PS Can we all get through this mess without insulting each other's intelligence? I get that we're all mad and scared at what is going on with the virus and economically. Let's continue to treat each other with mutual respect. I'm getting tired of being told that I must be stupid if I believe something.

A two-way street is required; mutual respect, one for each other. That rarely happens in these partisan times.

I wanted to use an example, but that would likely draw out the thread police... (TAKE IT TO THE POLITICS BOARD)...

But I would argue the modelers don't do that. Why? Because their hockey stick graphs are nothing if not consistent on slope.

The 2 million death prediction in the USA was the dumbest model in the history of infectious diseases, and learned people everywhere trusted it.

Bonfire1996 said:

---

But I would argue the modelers don't do that. Why? Because their hockey stick graphs are nothing if not consistent on slope.

The 2 million death prediction in the USA was the dumbest model in the history of infectious diseases, and learned people everywhere trusted it.

---

They are doing it. You may argue they're doing it poorly, and that's understandable.

In the beginning of this mess, the CDC presented the white house 4 scenarios based on their modeling. On the high end, 1.7MM people would die in this country if nothing changed about our behavior. It showed different scenarios down to 200K people dying if we did shelter at home orders, assuming that there would be some level of civil disobedience. This was done not to tell the President that 1.7mm people would die unless he ordered a nation wide lockdown. It was done in order to show a range of possibilities, GIVEN CERTAIN ASSUMPTIONS and what we knew at the time.

The Imperial College model similarly predicted 2.2MM in a do nothing scenario, but revised down their prediction to something closer to what we are predicting now.

In both cases, the "do nothing" scenario was AT THE TIME obviously academic in nature. It was meant to represent a worst case scenario, the high end of the possibilities, an unlikely one at that.

Again, I will circle back, I think that the models have actually performed pretty impressively. The fact that basically a month out IHME predicted nationally the curve of deaths to such high degree of accuracy is good.

Predicting no change in human behavior, and publishing that model's figures is nothing but a political stunt to attempt to prove their mental superiority. When in reality, it exposes that person as a rube.

There is no health event, no matter how minor, that involves massive death that doesn't force a change in human behavior through simple survival instincts. To act like we are ants who will walk through poison time and time again is just plain stupidity. It is so stupid, it must have an anterior motive, and it looks more and more like it is to establish fear and command obedience.

Bonfire1996 said:

---

Predicting no change in human behavior, and publishing that model's figures is nothing but a political stunt to attempt to prove their mental superiority. When in reality, it exposes that person as a rube.

There is no health event, no matter how minor, that involves massive death that doesn't force a change in human behavior through simple survival instincts. To act like we are ants who will walk through poison time and time again is just plain stupidity. It is so stupid, it must have an anterior motive, and it looks more and more like it is to establish fear and command obedience.

---

But the CDC didn't predict no change in human behavior. They just showed what would result if there was no change. They also showed 3 other cases that had various levels of human changes.

The CDC also did not release their model results publically, so I don't see why it was a political stunt. It was leaked to the NY Times who reported it. It was presented internally to aid in informing the WH and decision makers.

And again, getting back to insulting intelligence. I don't think these people are stupid when they try to show decision makers a range of possibilities. If people in the media try to take what was presented and twist it to make a scary narrative, they're the ones being manipulative to meet their own ends. However, the people making these models are just doing the best they can to inform decision makers. And again, their models have been really good considering all of the uncertainty that they have been dealing with, as the 538 article very artfully details.

Bonfire1996 said:

---

But I would argue the modelers don't do that. Why? Because their hockey stick graphs are nothing if not consistent on slope.

The 2 million death prediction in the USA was the dumbest model in the history of infectious diseases, and learned people everywhere trusted it.

---

Read the report. The 2 million deaths in the US was based on absolute do nothing - no contact tracing, no quarantine, no spontaneous social reaction. And, so, because they're not in fact stupid, they didn't "predict" this. Pretty much the entire premise of your OP and certainly this post is factually incorrect.

For example, Imperial noted in the report:

Quote:

---

...the impact of many of the NPIs detailed here depends critically on how people respond to their introduction, which is highly likely to vary between countries and even communities.

...it is highly likely that there would be significant spontaneous changes in population behaviour even in the absence of government-mandated interventions.
...
In the (unlikely) absence of any control measures or spontaneous changes in individual behaviour, we would expect a peak in mortality (daily deaths) to occur after approximately 3 months....

In total, in an unmitigated epidemic, we would predict approximately 510,000 deaths in GB and 2.2 million in the US, not accounting for the potential negative effects of health systems being overwhelmed on mortality.

---

It wasn't dumb, and it wasn't a prediction. It was a baseline. A baseline is not a prediction, it is something to measure against. Never-did-they-ever say "we think 2 million deaths are going to happen". And, never-did-they-ever "revise" that baseline down.

Case isolation and household quarantine of known cases by themselves reduce this by ~30%, in that same study - and both the UK and the US were already doing this.

The biggest problem here is a combination of ineptitude and shock-rating value in the media. Combine that with credulous pundits and media consumers, and you get really bad takes and opinions based in disinformation.

Given the consistent nature of epidemics, the unmitigated outcome is probably the most accurate model, not the least. That's what we'd expect if humans were a herd of animals, not prone to react. Everything goes from that baseline.

Why? The do-nothing case is the baseline to judge against. If the do-nothing case predicted 12,000 deaths over a year, would you see the value? Would you expect massive spontaneous reaction?

You don't like the unmitigated epidemic number because it's bad, or something. That doesn't make it un-useful.

"If anyone says if I jump out of a plane at 10,000 feet that I'll, die, I'm calling that person a dumbass. Everyone knows you need a parachute."

Ok - but what size? The study wasn't made to answer the question "what happens if we do nothing?" The study was made to answer the question - what are the effects of various interventions?

How do you measure those without the baseline do nothing?

k2aggie07 said:

---

Why? The do-nothing case is the baseline to judge against. If the do-nothing case predicted 12,000 deaths over a year, would you see the value? Would you expect massive spontaneous reaction?

You don't like the unmitigated epidemic number because it's bad, or something. That doesn't make it un-useful.

"If anyone says if I jump out of a plane at 10,000 feet that I'll, die, I'm calling that person a dumbass. Everyone knows you need a parachute."

Ok - but what size? The study wasn't made to answer the question "what happens if we do nothing?" The study was made to answer the question - what are the effects of various interventions?

How do you measure those without the baseline do nothing?

---

Because baseline do nothing models can be better. It can be non-linear. It can predict survival instincts. Baselines are worthless, especially when you readily admit the likelihood of a baseline is next to 0%.

Be better. That's it. Here's the deal, if a private sector prediction group made as bad a prediction as infectious disease modelers, they'd get exactly two chances to be as wrong as they have been throughout this ordeal.

Quote:

---

Because baseline do nothing models can be better. It can be non-linear. It can predict survival instincts. Baselines are worthless, especially when you readily admit the likelihood of a baseline is next to 0%.

Be better. That's it. Here's the deal, if a private sector prediction group made as bad a prediction as infectious disease modelers, they'd get exactly two chances to be as wrong as they have been throughout this ordeal.

---

You're missing the point of the baseline. The baseline with survival instincts still needs the baseline. Why? Because what causes panic? What causes spontaneous reaction? 1/1,000,000 deaths? 1 in 1,000? 1 in 100? How do you know when you'll hit that number?

The model didn't predict 2.2 million death as a plausible outcome, any more than if you ask me "what happens if I jump out of a plane at 10,000 ft without a parachute?" and I say "you'll die", that's making me predict your imminent death.

Have a better take, the model is useful. Your interpretation of it is wrong.

Belton Ag said:

---

Quote:

---

Read the report. The 2 million deaths in the US was based on absolute do nothing - no contact tracing, no quarantine, no spontaneous social reaction. And, so, because they're not in fact stupid, they didn't "predict" this. Pretty much the entire premise of your OP and certainly this post is factually incorrect.

---

As someone who actually doesn't doubt the 2 million death number if nothing changed, an argument could be made that the number is very likely not to be achievable. Even if governments, in defiance of everything we know, ordered no shelter in place and no distancing, humans, being rational actors in aggregate, would aldapt and distance themselves. They could host football games, but few would show up, etc.

---

This is my entire point, just with less dirty language.

The writers of the 2 million death baseline think very little of the human populace's general intelligence and survival instincts. And to me, it absolutely disqualifies them from any future modeling.

Quote:

---

The writers of the 2 million death baseline think very little of the human populace's general intelligence and survival instincts. And to me, it absolutely disqualifies them from any future modeling.

---

Except they don't, because they didn't. What should be disqualified is you interpreting models against what the authors of those models explicitly say.

And then telling them they're stupid or have an "anterior" motive.

Bonfire1996 said:

---

Belton Ag said:

---

Quote:

---

Read the report. The 2 million deaths in the US was based on absolute do nothing - no contact tracing, no quarantine, no spontaneous social reaction. And, so, because they're not in fact stupid, they didn't "predict" this. Pretty much the entire premise of your OP and certainly this post is factually incorrect.

---

As someone who actually doesn't doubt the 2 million death number if nothing changed, an argument could be made that the number is very likely not to be achievable. Even if governments, in defiance of everything we know, ordered no shelter in place and no distancing, humans, being rational actors in aggregate, would aldapt and distance themselves. They could host football games, but few would show up, etc.

---

This is my entire point, just with less dirty language.

The writers of the 2 million death baseline think very little of the human populace's general intelligence and survival instincts. And to me, it absolutely disqualifies them from any future modeling.

---

You act like all models are the same, with the same inputs and the same goal as far as predictions....

You saying that a model - that you are refusing to actually attempt to understand - is invalid and that the data scientists should be "disqualified" from future modeling would be about like me packing a messenger bag for a 2 week trip to Europe and then complaining that the manufacturer should never be allowed to make baggage again because it was so far off from what I needed.

Bonfire1996 said:

---

Taking current state and applying future projections.

I don't understand how learned people can take them seriously. Have you ever moved cities? When you go to the grocery store the first time and don't know where shlt is, it takes for effing ever. 38 weeks later you can visualize the same tortilla shelf with your eyes closed.

These are smart people building models. And they flipping suck at it.

Models don't take into account:

1. Self social distancing in fear
2. Medical advancement
3. Therapeutic remedies being discovered and tested
4. Public adoption of PPE
5. Other changes in population behavior.

These are obvious things to even the most average layperson. But every model forecasts a sheep like populace acting like the cast of idiocracy. Do the scientists think the world populace has exactly zero survival instincts? Because that's how they model. It's freaking asinine.

It's also why I think this fear of a second peak or second wave in Winter 2020 is dumb. They act like we aren't going to learn anything in 8 months. Who the eff thinks that? Think of what we have learned in just the last six weeks! Are we going to stop learning for the next 30 weeks?

Have faith in the American system. Really smart people have the opportunity to become really rich. They will figure it out sooner rather than later.

---

Have you met some of those academia folks?! They hacer ZERO common sense and self awareness.....and I mean ZERO!!

I think people lump "models" in together. IHME is not the same model as being used by Imperial. As far as I know, IHME doesn't use any population density, and is only matching curve for curve, using only this outbreak. Clearly this provides some bias toward expecting every new outbreak to behave like past ones, and since past ones have been largely urban...

The unmitigated epidemic curves from Imperial differ in both timing and peak height, and I assume the population density drives that.

Bonfire1996 said:

---

Taking current state and applying future projections.

I don't understand how learned people can take them seriously. Have you ever moved cities? When you go to the grocery store the first time and don't know where shlt is, it takes for effing ever. 38 weeks later you can visualize the same tortilla shelf with your eyes closed.

These are smart people building models. And they flipping suck at it.

Models don't take into account:

1. Self social distancing in fear
2. Medical advancement
3. Therapeutic remedies being discovered and tested
4. Public adoption of PPE
5. Other changes in population behavior.

These are obvious things to even the most average layperson. But every model forecasts a sheep like populace acting like the cast of idiocracy. Do the scientists think the world populace has exactly zero survival instincts? Because that's how they model. It's freaking asinine.

It's also why I think this fear of a second peak or second wave in Winter 2020 is dumb. They act like we aren't going to learn anything in 8 months. Who the eff thinks that? Think of what we have learned in just the last six weeks! Are we going to stop learning for the next 30 weeks?

Have faith in the American system. Really smart people have the opportunity to become really rich. They will figure it out sooner rather than later.

---

I wrote my own model and attempted to model all of that. Do some internet sleuthing and you will see others tried to model that and more. But the models are still woefully inaccurate due to the lack of specific knowledge about the virus and the high degree of variability in social interaction.

I think the model haters are acting like the sheep you refer to. They all recite the same tropes and would rather let their intuition guide their actions than apply some science.

I am not saying models should lock in our plans, but doing nothing is not wise either. I say model it as best you can and use real data to constantly adjust the model and make decisions.

Models with so many high stochastic errors are subject to manipulation and confusion even by those that know better.