From the study proving that Neil Ferguson’s Chinkvirus model contained flawed methodology (to say the least) comes this conclusion:
“On a personal level, I’d go further and suggest that all academic epidemiology be defunded. This sort of work is best done by the insurance sector. Insurers employ modellers and data scientists, but also employ managers whose job is to decide whether a model is accurate enough for real world usage and professional software engineers to ensure model software is properly tested, understandable and so on. Academic efforts don’t have these people, and the results speak for themselves.”
Hell, considering what’s come out of academia in terms of climate modeling as well as this latest fiasco, I’d prefer to have bookies produce the models, rather than universities.
And this is why charlatans like the Hockey-Stick guy (of global warming infamy) steadfastly refuse to release their code — they know it’ll fall over under the slightest scrutiny.
Had I ever tried to get this bullshit past my clients back in the day when I was involved in this kind of thing, I’d have been fired on my ass and my business cred utterly demolished. These pricks deserve no less.
I think Ferguson is a damn good historian overall. But he gets over his ski’s with this kind of crap and he’s done it before making it hard to defend some of his better work. He did something similar with Brexit IIRC. This kind of thing infuriates me because of the very thing the article writer states, Academia is an isolated cocoon that has gotten so much mileage out of their authoritative fallacies never being challenged, that they are faced with a huge credibility crash, for positions they absolutely cannot defend. Then they wonder why anyone intelligent enough to wipe their own ass, and not part of the fraternity, doesn’t trust them and they get shrill, and start screeching “Racism” or “Denier” at anyone.
The trick with this stuff is the “Findings” are published for peer review, but not the models or assumptions that created the findings. Its like James Randi once said. Scientists are easy to fool, because they don’t realize the trick was done two days before the experiment.
I had a conversation with a published PHD historian online (or someone who claimed to be) recently who absolutely was incapable of defending his thesis. Even with a link to his book. All he could do was say racism. I suspect that this has worked so many times with his colleagues in the closeted fever swamps, its become the modern equivalent of making “the sign of the cross” for whatever purpose, a ritual completely devoid of meaning and rigor. My ultimate conclusion was he didn’t defend his thesis because he COULDN’T defend it, and had never had to. (BTW he is the one who started the argument, and I merely asked him to link to his published work, or give me a summary of his thesis”). Unfortunately its an hour of my life that is gone forever. It’s easy to just say move on with your life but this guy is going to “teach” someone someday.
Long and rambling, but I had to get it off my chest.
/Rant
I believe you are confusing Niall Ferguson, Scottish historian, with Neil Ferguson, British Epidemiologist.
This has been happening a lot lately, you’re not alone.
Cheers,
JC
By The Lord God Almighty, I am. I ashamedly retract that element of it. Though I still believe my overall comments are sound.
Thankee kindly for the correction.
Oh, yes, great comments! Don’t feel bad, the names are so close it’s happening a lot lately. Niall has been famous a lot longer than Neil, and it’s causing quite a bit of confusion.
Cheers!
Fixed, JC, thankee.
[/1984]
Fucking Celtic/Gaelic names will be the death of me yet.
Like you, Kim, I have some experience with this kind of stochastic modeling in finance. Some 20-25 years ago something in finance called Portfolio Allocation Optimization Software was all the rage. I remember a PhD from U Penn doing a session on using a software optimizer at an investment conference back then. I won’t bore you with the mathematics behind it, but these suffered from the same things the Imperial College model suffered from, apparently.
Long story short, the prof did a sensitivity analysis by changing just one input–long term stock market returns– by 2% average annual return. EVERY other input he left the same. With the higher figure the optimizer suggested a portfolio of 80% stocks and 20% bonds, and with the lower figure it suggested a portfolio of 20% stocks and 80% bonds. Because these widely disparate results were obviously ridiculous, it was common practice to put upper and lower bounds on the allowable OUTPUT, believe it or not. SMH.
I raised my hand during Q&A and asked the prof if he predicted, say, an 8% stock market return over the next ten years, and it actually came in between 7% and 9%, he’d be a pretty incredible market predictor, would he not? He enthusiastically agreed. Then I asked him, “Well if that’s true and the difference between those two assumed future returns yields a 60% swing in the stock allocation output, why the fuck would we spend a couple thousand dollars on one of these programs? Why not jut lay out several different periods of historical stock market performance on a legal pad for the client and then tell the client to stick with the ‘traditional’ 60/40 allocation and take advantage of big bull markets by trimming allocation a bit afterwards, and of bear markets by leaning into stocks afterwards?
He was clearly flulmmoxed by my question, but a lot of investment advisor heads were nodding in the audience. That software fad lasted about 2-3 years, and you don’t see much of it any more. Fucking University eggheads are fine until you start to give them real world stuff to play with.
Anyway, I agree with you, let actuaries work on this stuff, they’d be much better. By the way the Federal Reserve is chock-a-block full of arrogant PhD idiots like this, too. If you have the time and inclination during the lockdown I highly recommend the book, Fed Up, by Danielle DiMartino Booth, where she goes through how all the Fed assholes kept referring to their precious models during the 2008-2009 market meltdown. “That may be happening in the real world, but it will never work in theory!” Same thing with this Shamdemic, I think.
Cheers,
JC
Folks, software engineer/architect here. I’ve looked at that code repo. Utter trash doesn’t begin to describe it. It flat out nauseates me as a professional.
I took a peek too, but I laughed because I spent years as a “Maintenance Programmer” and I’ve seen much worse.
If anyone else wants the joy of reading this awful stuff, it’s here – https://github.com/mrc-ide/covid-sim.
The main file is CovidSim.cpp, 8209 lines of computational gibberish.
https://github.com/mrc-ide/covid-sim/issues/168
This particular Charlie Foxtrot was fixed yesterday.