From the “all models are wrong, some are completely useless” department. This was originally from May 6th, but it’s so bad, we deserve a reminder just as we are about to emerge from the final phases of lockdown here in California.
From the National Review:
‘Professor Lockdown’ Modeler Resigns in Disgrace
Neil Ferguson is the British academic who created the infamous Imperial College model that warned Boris Johnson that, without an immediate lockdown, the coronavirus would cause 500,000 deaths and swamp the National Health Service.
Johnson’s government promptly abandoned its Sweden-like “social distancing” approach, and Ferguson’s model also influenced the U.S. to make lockdown moves with its shocking prediction of over two million Americans dead.
Johan Giesecke, the former chief scientist for the European Center for Disease Control and Prevention, has called Ferguson’s model “the most influential scientific paper” in memory. He also says it was, sadly, “one of the most wrong.”
Full story here
I wonder how long it will take for people to realize that some climate models like RCP8.5 and the upcoming “hotter” IPCC models will be equally useless and damaging?
I started programming for hire in third year university. My first program worked well, and saved us countless hours of manual calculations. This was long ago, in 1970. The computer was an IBM360 – the languages were Waterloo Fortran, aka WATFOR and FORGO (earlier more primitive languages were ONCEUPONATRAN AND BEFORGO).
I subsequently programed in GPSS, a simulation language, ALGOL and other languages. and much later produced very large scientific, engineering and economic models in Lotus123 and later Excel. My models worked well because I really sweated the details, and checked and re-checked them – hindcasting and forecasting. I think it requires great diligence to produce an accurate model.
I’ve also found some egregious errors in the models of others, even models that had been used for years for submissions to government regulatory agencies. Those models produced total cr@p.
Some modelers bash together a model and if it produces results, they’re done! Doesn’t matter if the results make no sense, don’t hindcast or forecast correctly, or essentially produce nonsense – it’s done by computer – it must be right! Right? … ah, No – not even close! Climate models fall into this category – they don’t even have the basic physics correct – they just produce total cr@p – and yet our idiot politicians have squandered many trillions of dollars of scarce global resources to “fight runaway global warming”, aka “fight climate change” – the first of which is NOT happening, and the second has happened since the dawn of time.
What is even more foolish is the idea that they can combine the results of dozens of cr@p climate models and produce a better result – ask any little kid what happens if you combine many small piles of cr@p – you just get a much bigger pile of cr@p! And these modelers were paid to produce this nonsense, which is why they do it – if they admitted it was all cr@p, they’d all be out of work!
The mainstream computer climate modellers have made about 50 very-scary climate predictions, and every one of them has failed to materialize – nobody should believe them.
We’ve done a lot better, as follows:
https://wattsupwiththat.com/2020/06/03/alarmist-queen-hayhoe-takedown-by-friends-of-science/#comment-3008219
Told you so – 18 years ago. [That includes you, Michael Moore.]
The ability to predict is the best objective measure of scientific and technical competence. Every very-scary prediction of runaway global warming and climate chaos made by the global warming alarmists has failed to materialize. Nobody should believe them – about anything.*
Following are our two major statements we published in 2002 – these statements are correct-to-date, for anyone who understands climate and energy. The climate alarmists and their slave leftist media, with their “100% wrong predictive track record”, will dispute them. *See note above. 🙂
Regards, Allan MacRae
_________________
OUR TWO MAJOR STATEMENTS PUBLISHED IN 2002
In 2002 co-authors Dr Sallie Baliunas, Astrophysicist, Harvard-Smithsonian, Dr Tim Patterson, Paleoclimatologist, Carleton U, Ottawa and Allan MacRae wrote:
1. “Climate science does not support the theory of catastrophic human-made global warming – the alleged warming crisis does not exist.”
2. “The ultimate agenda of pro-Kyoto advocates is to eliminate fossil fuels, but this would result in a catastrophic shortfall in global energy supply – the wasteful, inefficient energy solutions proposed by Kyoto advocates simply cannot replace fossil fuels.”
Edit:
The mainstream computer climate modelers and their minions have made about 50 very-scary climate predictions, and every one of them has failed to materialize – nobody should believe them.
Never, ever trust pencil-neck marxist modelers.
The jerks that listened to this guy ruined the world economy.
When the dust settles, serious studies must be done with all the parameters entering the epidemiology study of a rapidly spreading infection, the health authorities of each country should settle on a way to face it in the future, assuming COVID 19 was a test. We were lucky the infection was not more virulent than the flu. Next time one with a death rate closer to the medieval plague could spring up, which left some areas with 50% dead, not just the old and infirm.
In most countries the real choice was not between number of deaths and the economy, although in the news it sounds like that. It was between the saving of the health care system , avoid its demolition as was clearly seen in what happened in Lombardy. I think those videos of dead in the hospital corridors and trucks carrying coffins influenced the government decisions more than any models .
Nobody has answered the question: “can an economy survive if the health care system is demolished”?
In retrospect each country need only do lockdowns in specific regions and facilities, which they still do in Greece, even now when general lockdown has been lifted . Certainly the damage to the economy would have been less. Hind site is useful only for planning for future reactions.
What evidence exists that the lockdown was not the cause of most contaminations?
The problem is not that Neil Ferguson was wrong; the problem was that it took the politicians in the US and UK too long to wake up to the threat of Covid. His prediction of 500,000 deaths in the UK didn’t need a computer model; you could do it on the back of an envelope. The death rate from Covid is 1%. That was known in February. The UK population is 66m. 80% infection = 500,000 deaths. Q.E.D.
Despite lockdown we’ve had the worst death-rate in the world here in the UK (40,000 recorded so far, over 60,000 excess deaths, heading towards 100,000 – so Ferguson wasn’t wrong! He was late.). Why is it so bad here? Because we delayed too long. You either go into lockdown early or you don’t bother (NYC probably shouldn’t have bothered in the end > 25% infection rate anyway so lockdown probably only halved the death rate). Every week you delay lockdown increases the death rate by a factor of ten and doubles the time you need to be under lockdown to get the infection under control. This is obvious from the daily infection rate curve. That is the failure.
As for the modelling: the problem is that epidemiologists are like climate scientists – 2nd rate academics who peddle mathematics (or physics) they don’t understand. The maths of infection is not difficult. It is 12th grade stuff. It is the same as the maths of nuclear chain reactions (1st order differential equations). Moderators in nuclear reactors perform the same function as social distancing does in epidemics. It reduces the R number. So if you want someone to model your outbreak, ask a nuclear physicist, not a medic or social scientist.
Can you provide evidence for any of that?
The Model E code is almost as amateur as this disaster. I’ve looked.
Some facts – not all code libraries use the same pseudo random number generators so same seed gives different results on different systems. – not all CPU’s use the same ‘floating point’ processing algorithms – error propagation via inherent rounding will always taint a model like this – stochastic (ie, iterative with random numbers) is wrong, he should have the mathematical skills to create the partial differential equations that represent the ‘true’ model, not a lazy monte christo simulation…
The climate is a chaotic system. It cannot be modeled well for long-term forecasts.
The weather is a chaotic system. It cannot be modeled well for long-term forecasts.
The economy is a chaotic system. It cannot be modeled well for long-term forecasts.
Virus propagation follows known math. It can be modeled well.
Days from infection to symptoms plays a major role in early models when data is sparse. When days-to-infection is close to zero quarantining the ill works. Larger numbers indicate masks and distance from possible carriers. A large number, like 5, indicates a steep initial curve. When the days-to-double rate is known through data one must change the model. Time from infection to symptoms is merely interesting. Time to try a Gompertz function.
When years and years of data (like, I dunno, HCQ) are available no model is necessary. Clinical data: It is safe. Hippocratic Oath satisfied. CAN’T HURT; might help if enough Zn in patient. Forgive off-topic rant.