The Model Was Incorrect
It seems that the “experts” predicting massive die-offs using “Ferguson’s Model” and insisting that we all impinge our lives with lockdowns and social distancing were no experts. Instead they are a bunch of amateurs pretending to be programmers when they are not professional programmers, and every bit of code that they used was infested with bugs that not only throw off the results, but can’t repeat them. Their response to this error is to tell people to “take an average”, which is incorrect. Their results were spectacular, grabbed news media’s attention and were wrapped around the world like a rope.
But the spectacular always gets attention, which keeps repeating itself until the curtain is dropped and the reality of just plain bad, error-filled work raises its head.
So the entire world gets put into pandemic panic and goes into shutdown when it wasn’t really necessary. And no one has asked why we don’t do the same things during flu season, when flu of any strain is just as deadly. Yes, I did wonder about the extreme measures that have been taken to prevent the spread of this bug, which has not worked, and why it’s been given so much attention when the flu can be just as lethal, but – well, we all get flu shots, don’t we? And haven’t we forgotten the panicky reports over deadly swine flu a few years ago, which seems to have been forgotten? There was the Hanta bug a few years back, too, which was prevalent in the southwestern USA, and was finally traced to mice as carriers.
According to the author of this article, a professional programmer with several decades of experience at it, their efforts have resulted in both inaccuracy and a cavalier attitude toward their inconsistent results, which is a dangerous way to behave. When software is so poorly done that multiple inputs of the same information produce a wide variety of results, there is something wrong with the people who created the coding in the first place.
Here’s the link to the professional programmer’s article:
https://lockdownsceptics.org/code-review-of-fergusons-model/
Read the whole thing. She doesn’t waste words anywhere.
Category: "The Floggings Will Continue Until Morale Improves", COVID-19
Neil Ferguson not available for comment, as he is busy dorking his left wing married lover while spreading his chicom flu to all the folks of lower London and beyond…
I was just about to ask if he wasn’t the one who just got caught violating his own rules to tap his married-with-kids side piece. And right after much was made of him testing positive for Chinese Bat Flu himself.
What a fucking douchebag.
Frau STAATS is the first hot socialist woman since Eva BRAUN.
Honey traps are a commie specialty.
Does that make her “StatsPolezi” ?
My deepest sympathies to all of the people affected by this ChiCom Originated Virus Infecting Disease of 2019 (COVID19). At this time Randolph Hearst is dancing with joy in hell and chortling with glee. He is anxiously awaiting the arrival of Rahm Emmanuel so that they can celebrate the use of yellow journalism to ensure that a crisis is never allowed to go to waste.
Just as the issue of gunz control is not the control of the gunz as much as it is the control of the people, thus we have seen that the control of the people is more important than the control of the disease. The dictatorial governators have seen how they can expand their powers over the people and some of the deep rooted domestic enemies have seen how they can pull the wool over the eyes of elected officials that have to trust them to give them only the facts, not a skewed version of what they want them to be.
Our freedoms have taken a bigger hit than the economy.
“All models lie. Some are useful.”
M. Guyotte
Of course they were wrong. Models usually are. They are most useful in creating general plans until enough actual data is available to make specific plans. In other words, models may help you start moving things in the right direction until an exact situation can be determined.
While the scammers indeed are very bad guys, those who fall for their disinformation bear significant responsibility for purposefully causing great harm to an entire country using/unproven faulty data. Basing all this misery on trusting what unknown entities are telling you is just unconscionable.
Bullshit, Ex,
The initial estimates of 1 million to 2 million were based on the US doing nothing to stop or slow the spread.
We acted and we flattened the curve.
More than a thousand people are still dying per day but we no longer have exponential growth; due to the lockdowns.
And this is not even close to over yet. This is just the first 6 months of a pandemic that likely will last 18 months.
All the peer reviewed models show a return to exponential growth if the lockdowns end.
The Trump ad insist rational n is showing some governors a model that indicates the growth will not increase if lockdowns are ended. The administration refuses to release the model to the public and refuses to release details of the model. So it has faced no scrutinize from statisticians, scientists, or epidemiologists.
I have no idea why you keep feeling the need to spread the bullshit coming out of the White House. It has been continuously debunked, often by Trumps own experts, and Trump is putting this information out for purely political reasons with no regard to public health.
He considers the pandemic a PR problem, not a public health problem. For him these press conferences are campaign events.
Aim also says a lot that you claim to believe this shutdown was not necessary but you have no intention of doing anything to put yourself at risk.
You want others to expose themselves to the virus for economic reasons.
But not you. You are going to continue to social distance and continue to make sure you are not exposed.
Because you know damn well that your are In a very high risk group for life threatening complications if you catch the virus…
And you care fuck all about anyone else.
“Flattening the curve” does not reduce deaths by infection. It slows down the rate of infection spread so that the hospital capacity is not overwhelmed. Mathematically speaking , the area under both curves is equal.
You appear want the current panic-induced mayhem to consider for political reasons. Shame on -you-.
I want folks to go back to living their lives without these bullshit restrictions because Liberty is -way- more important that safety. Even my safety.
Choosing to self isolate is a great thing if you feel it’s needed. That’s far different from government mandating the Un-Constitutional suspension of our First Amendment rights (Freedom of assembly), and destroying businesses and our supply chains in the process. As 11-B points out, “flattening the curve” does NOT change the numbers, only the rate at which they occur. More and more info is coming out showing that the virus was likely here months before we thought. Demanding that these ridiculous lockdowns end does not show callousness, but concern for the population, as hospitals are furloughing staff, and may quite possibly go bankrupt due to lack of income because all “ellective” procedures have been stopped. In addition, we’re seeing large spikes in suicides, child and domestic abuse, drug and alcohol abuse, and deaths from other medical conditions that people are not seeking help for because they are not allowed, or they’ve been too scared by the left. These are real cases, and deaths that are completely avoidable if we stop this silly panic, and get back to normal. The number of confirmed cases in Colorado is roughly 18,000, with the assumed number (based on over 80% of cases being asymptomatic) being closer to 90,000. With total deaths running at just over 800, (a number that is inflated, as anyone that dies while positive / with symptoms is being counted, regardless of actual cause of death, and that’s not an arguable point, I got it directly from the state officials responsible for such reporting), which give a mortality rate of less that 1%, putting it squarely on a par with the flu. While some of the initial reaction to this might be understandable, the religious like furor with which the left clings to this overreaction defies all sense and logic, UNLESS the actual reason behind it is to damage the country, it’s institutions, and it’s President (as some on the left have openly admitted). With some of these tyrant governors extending the lockdown until July 4th (you quarantine SICK people, when you add healthy people to the mix, it’s a lock… Read more »
Give prior health issues, if I catch a pneumonia I am probably a goner.
Open everything back up, right the fuck now. No kid should go hungry because alleged “grownups” panicked over what for most is a flu. No one should lose their fucking livelihood because shitbirds dont like who won the last election, and “crash the economy” is how they mask their incredibly failed and corrupt methods for picking candidates.
Open it all back up. Mock and ridicule those who continue to bang the panic drums.
Agree 100%. No-one is telling anyone that they HAVE to go out, they’d be just saying you CAN…
Here you go, Lars… Cuomo and his policies have left him and his minions with blood on their hands:
https://libertyunyielding.com/2020/05/08/media-darling-cuomo-mismanaged-covid-19-killing-thousands/
Yeah.
We are into “when is the tribunal?” territory.
Another Proggie, running up the body count.
You can add de Blasio to the list. NYC is responsible for the spread of 60 to 65% of the covid19 cases across the country.
https://www.nytimes.com/2020/05/07/us/new-york-city-coronavirus-outbreak.html
“And you care fuck all about anyone else.”
This is absolutely hysterical coming from the asshole who is thankful for the deaths of older American citizens. I’m sure the irony escapes you.
That is not what I said. It is what some of you chose to interpret.
Liar.
You declared that the disease was expected to be disproportionately lethal to a particular demographic, expressed your dislike of said demographic, and called it a good thing. Not a whole lot of room for interpretation in that one, dickead.
That’s one of the biggest problems everyone has with you, Lars. It’s not just that you say dumb, appalling shit. It’s that you say it and then try to weasel your way out of it when called out, at times even denying that you said it when your exact words are visible on the very same page. It’s a particularly petty form of cowardice that strips every shred of credibility you could ever hope to claim.
BINGO!
Amen!
Second.
Koalemos (AKA Commissar AKA Poodle AKA Seagull AKA Cthulhu): That is EXACTLY what you said, you mendacious twit. You directly stated that you were thankful that the Wuhan Coronavirus would target older people because they were more likely to oppose one of your favorite political goals: single-payer, government-managed healthcare. What follows is a direct quote from one of your previous comments. I’ve added emphasis. I am thankful it will impact the generation most to blame politically for not letting our country prepare of it. Lest you forget: the “it” in the quote above refers to the Wuhan coronavirus. Elsewhere in comments, you identified the elderly as that generation most likely to opposed nationalized healthcare. The context of the discussion also highlighted the fact that the Wuhan coronavirus is disproportionately deadly to those over 65. Given the those additional facts and context, anyone who has a reasonable command of standard English can understand your actual meaning. The first portion of that sentence clearly says that its author is thankful that the Wuhan coronavirus is disproportionately lethal to the elderly. Further, the italicized portion clearly identifies your motivation for that ghoulish thanks. You didn’t say it was Karmic rertribution. You didn’t say it was a case of “what goes around comes around”. You said, clearly, that you were thankful for what was likely going to happen – and also indicated the reason for your thanks (specifically, because it would happen to those who opposed one of your political dreams). Your words. They’re not negated by the semi-obligatory “I don’t want them to die” boilerplate you included at the beginning of the original comment. Including that merely makes you look disingenuous or a fool – because it means you’re either being insincere or contradicting yourself. You’re effectively saying, “I don’t want them to die, but since they opposed me politically I’m glad they will.” Your words. Own them, or disavow them. But don’t waste our time claiming you never said them, or that they mean something besides what they clearly say to anyone who can read and comprehend standard English. Oh, and here’s… Read more »
Your rage is your undoing. Repent.
Relocate.
The lockdown was completely unnecessary if, as suggested by recent evidence, the China virus was present in the US as early as December of last year, as many of us suspected.
https://www.yahoo.com/news/evidence-suggests-coronavirus-likely-spreading-154500955.html
Yeah, December, meaning 6 months ago. We already knew that.
Geneticists were reporting that back in February.
I have no idea why the press acted like it was a knew revelation.
If we did not lock down the exponential growth in cases would have continued and the exponential growth in deaths would have continued. Some communities would have had cases exceed medical capacity and many would have died that might have been saveable if we had not be overwhelmed by sales.
The lockdown was necessary.
You choose to believe it wasn’t for purely political reasons.
All of the science consistently shows the lockdown was necessary. He only people claiming otherwise are partisans and politicians.
This thing is not over, and lockdowns are being lifted in some areas soon.
So you will get to see how moronic you are in real time.
“All of the science consistently shows the lockdown was necessary. He only people claiming otherwise are partisans and politicians.”…
Tell that to Sweden
Give it a rest, dipshit. If true, the virus was spreading across the nation for at least three months prior to the lockdown. Three months. Let that soak into your pea brain for a moment. Only a truly moronic person would think that the spread of the virus was able to be controlled three months after it was spread across the nation. Exponential growth had already occurred, imbecile. And the science is only as good as the data at any given time. It changes as new data is acquired. So tell us, master of the moronic, what does the science have to say regarding this “new” revelation? Of course, this thing isn’t over yet! How long do we stay in lockdown? How much more damage needs to be done to the country, the economy, and our rights before dumbasses such as yourself think it’s “safe?” I know! Go ask your Communist friends in China what they think. Maybe you can get them to tell us the truth. Just show them all your ’empirical data.’
“How much more damage needs to be done to the country, the economy, and our rights before dumbasses such as yourself think it’s “safe?””… That’s an easy one!
Until the left thinks they can install themselves as “supreme leader” and institute the tyranny they’ve been giving us all a taste of!
On a related note, as restaurant here is defying a “cease and desist order” and says they’ll take it to the Supreme Court… This needs to start happening across the country, as it appears to be the only way short of bloodshed to end this crap
Fyr… check out the links below. Nursing homes = death camps in libtard states.
On a related point, if these lockdowns are so freaking successful, how is it that my Aunt and Uncle, who never left the house since this crap started, (and the rest of the family went out only when they had to) died from it???
Go play with your Empirical Evidence!
You really have nothing to add to any conversation do you infidel?
You rate nothing better Komrade Kazoo.
And we have no damn clue how many people have died worldwide. Most governments are lying about their numbers. And nearly half of the world’s governments utterly lack the capacity to even track something like this.
Which is the same thing that happened during the 1918 pandemic.
The lying of foreign states is no reason to scrap our essential Liberty.
Nobody is scrapping it. Lockdowns have a historical precedent in the US and things were far more rigidly enforced during the 1918 pandemic,
They even shot a few people that refused to comply during the 1918 pandemic.
This lockdown is more permissive and the consequences for breaking it are far less severe, other than the risk you put on yourself and others.
Your constitutional rights remain intact. You never had the right to endanger the health, safety, and lives of other citizens.
Quarantines are not for the healthy, shitbird.
Take off the jackboots and quit the panic-spazz. You look stupid.
“Lockdowns have a historical precedent in the US ”
Show me.
Americans of Japanese descent in relocation camps. And what party did that?
Hmmmm. Yeah, I guess I would classify that as a lockdown. Thanks for the correction.
Caw! Caw!–Gulp!–Ummm, tasty.
Eat more crow, as the cows say.
Really? Those arrested in various locations for trying to exercise their rights guaranteed by the First Amendment on Easter Sunday might tend to disagree with you, Koalemos (AKA Commissar AKA Poodle AKA Seagull AKA Cthulhu).
Amen !!!!
China lied to the world about this, Lars. How many people could have been saved if they were honest about this when they first started seeing it? Acting on this info earlier could have “flattened the curve” a Hell of a lot earlier and possibly reduced M&M (morbidity and mortality).
China lied about SARS… and they lied about COVID.
It is true they lied. But the consequences of those lies really primarily killed the Chinese people.
The WHO only believed China for about 4 weeks. By the 21st of January they knew China was lying and reported that to world governments.
US intelligence knew China was lying prior to that.
Trump chose to ignore the WHO, the CDC, and the intelligence community.
While he imposed a very performative and ineffective travel ban at the end of January he consistently reported that there was nothing to worry about even claiming the “15” us cases would be “0” soon, despite the fact that the confirmed cases were already several times more than 15 when he said it and despite the fact that est Mats at the time from the CDC was that they were only detecting 8% to 16% of cases.
He tried to handle the pandemic as a PR problem from the beginning,
That is what caused this to get out of hand.
Though it likely would have eventually since even after it is clear that this thing is a threat we still have more than 30% of the population insisting we have nothing to worry about,
You know Lars I realize that your whole prog worldview relies on massive deference to “top men”, but given how often our intelligence services have been catastrophically wrong about critical issues it’s no surprise that so many of us are not especially inclined to pay them much heed.
So maybe they were right about China this time, and maybe if they hadn’t pissed away most of their credibility for 40 years or so more of us might have been willing to give them the benefit of the doubt. But the erosion of trust in our institutions is largely due to the failures of those same “top men”, Ivy league douchebags almost to a man, who have so often been self-serving and deceitful in how they presented information to the American people and our leaders. I mean, I just saw Max Baucus, former ambassador to China, just compare the U.S. to 1930s Germany in how we deal with China while the Chinese government he was comparing us unfavorably to maintains actual concentration camps for Uighurs.
None of that excuses your girl Nancy P from telling people to come on out to San Francisco’s Chinatown in February, or the New York city health commissioner doing the same at the same time. None of it excuses De Blassio’s complete failure of leadership in New York, which accounted for about 1/3 of U.S. deaths recently.
You want to throw shade at Trump go nuts, but none of that explains away the fact that the vast majority of Covid deaths come from your true blue urban shitholes. Most of us out here in big red rural America feel bad for you, but we aren’t really surprised, cities can be wonderful places but they have always been dangerous as well.
“None of it excuses De Blassio’s complete failure of leadership in New York, which accounted for about 1/3 of U.S. deaths recently.”.. Actually, with the new reports that 60-65% of cases in the US came out of NY, it’s actually more like 2/3 s are on his hands…
“While he imposed a very performative and ineffective travel ban at the end of January…”
You are such a partisan hack. Would Trump’s actions have been similar to leftists such as Pelosi, de Blasio, and his health commissioner who were telling people everything was alright in January and to go eat in Chinatown? Maybe Trump’s “there was nothing to worry about” was based on the lies and misinformation your Chinese Communist brethren were feeding to the world via the WHO well into February and even today?
And how about that ineffective travel ban? Couldn’t have been ineffective because, as recent evidence points out, the virus had been spreading across the nation in December, perhaps even earlier?
Chinese Communist lies. That is what caused this to get out of hand.
“But the consequences of those lies really primarily killed the Chinese people.”
Really? So, you discount all the American deaths, especially those you’re thankful for, as well as the deaths all over the world? Those deaths are all attributable to the lies your ChiCom buddies told the world.
Yeah, we know, “Orange Man Bad!!!”
Fucking clown.
Oh, yeah! I almost forgot all the deaths Cuomo is responsible for by forcing New York nursing homes to accept Wuhan virus patients.
As I recall, the POTUS was soundly criticized by those on the left for implementing that ban at the time – which was just a few days after China started implementing quarantines in Hubei province, a few days after the WHO blew the whistle on China’s cover-up of the situation’s seriousness, and when only a handful of cases had been confirmed in other nations (representative sample from John Hopkins for 2 Feb: 1 case in France, 2 in Italy and Spain, 5 in the UK and France and 6 in Germany).
The EU also raised hell when we implemented travel restrictions on their nations in mid-March, too – though they all did the same not terribly long thereafter.
My point: the US response was both early and reasonable, given what was known at the time – and would have occurred much earlier had China not deliberately hidden the magnitude of the problem for six to eight weeks. Had they come clean back in in November (when the outbreak was well-known in Hubei), it might have remained a local issue instead of becoming a global pandemic.
OK, Koalemos (AKA Commissar AKA Poodle AKA Seagull AKA Cthulhu) – convince me. Just precisely how do you know this? Not suspect – know.
Did someone in the IC tell you – thus disclosing information for which you have no official need to know? Or are you simply (as you often do) assuming this or MSU (e.g., Making Stuff Up)?
Further: just when did the US IC know? If you have personal knowledge that what you claim above is true, you should also know roughly when the US IC knew the above unambiguously.
And remember: “suspect” does not equate to “know”.
I’m not surprised (despite Lars’ bowel eruptions, above) – because I watched how the predictions changed, knew that the input data was suspect from the get-go due to the Chi-Coms ‘face-saving’ efforts, etc.
I worked as a programmer for more than 25 years. I know that GIGO is very real.
I also know that programmers are made to re-run statistical reports until the politician demanding the report sees what he/she wants to see.
I also know that non-programmers who think they can do these things generate a beaucoup of garbage which they, in their delusions, believe Mean Something.
And Lars is right there with them.
It would be nice if Lars the Squabbling Angry Seagull had actually read the report at the link provided, wouldn’t it?
Since it was written by a REAL programmer, and debunked the results by the amateurs who put that crap together, one would think that carries weight. And it does, unless your sole agenda like Commissar’s is to cause disruption and try to insult people just to get a response.
What makes it worse is that Ferguson had Boris Johnson’s “ear” and took advantage of it, and everything just went from there. Meanwhile, Ferguson was playing pattycake with his married girlfriend.
You mean the article posted under a pseudo name? The one where there is no evidence the author has any relevant expertise? Certainly is not a epidemiologist.
On a site that is entirely created to make a political argument about the issue.
And the people who built these models being used by the WHO and epidemiologists are actual scientists. With decades of experience, PhDs, peer reviewed results, and have no agenda other than to give the best scientifically supported advice they can to save lives.
I have no idea why you think being a “programmer” gives anyone expertise in this. Coding schools are 6-12 weeks long. You can self teach using online sources, and YouTube.
The Bay Area churns our thousands a year and they sit in coffee shops and Panera bread, or work from home doing gig economy jobs.
Being a programmer is the 21 century version of the 1990s coffee shop dwelling “creative writer”
Not amateurs as you try to claim.
Why are you touting the WHO as a creditable source/agency?
” Certainly is not a epidemiologist.”
Funny you should bring that up. Neither is “Dr.” Ferguson. He is(?) a physicist who seems to enjoy playing epidemiologist but whose track record as such is ludicrous.
“Coding schools are 6-12 weeks long. You can self teach using online sources, and YouTube.”
LOL. Obviously the STEM curriculum was not for you. There is some truth, amazingly, in your last two or three paragraphs; there are a lot of posers out there who think some sort of “credential” is enough to give them some sort of expertise.
The author published under a pseudo name? Does he have a pseudo job? Maybe he has a pseudo house, pseudo wife, pseudo children, and even a pseudo dog – unless of course, he’s a pseudo cat pseudo person? And you claim to have a college degree in what? Dear God in heaven! Even a semi-edumacated dolt such as myself knows what a pseudonym is.
I read it – there are some good points, and some bad ones. Yes, the code is ugly, like most academic code.. and like most code used in insurance models too, at least the ones I’ve seen. A better approach than the laughable, “Let insurers do it!” would be to encourage open-source models so the bugs can be caught, and fund professional programmers to work on such things, not academics trying to publish-or-perish.
As for determinism, I think the author -whose background is in databases and non-numerical computing- confuses a few things, or perhaps that’s my misunderstanding from glancing through it quickly. But in any floating-point calculation, you have a few factors that render strict determinism difficult – chaotic effects from truncated precision being one, but another is that things like the commutative property in math doesn’t apply in computers.
In short, “a * b * c” should be doable as either “(a * b) * c” or “a * (b * c)”, with equal results… but it doesn’t. So build a model with the same code using two different build flags, and you’ll get different results.
This is why statistical ensembles of models are a necessity. This doesn’t excuse the bugs, but they aren’t ‘hiding’ results behind statistical results.. they’re avoiding -or hiding behind, if you prefer- the one thing worse than a buggy model: a model giving a singular answer that people take as gospel.
I will also, while not vouching for the validity of the Ferguson model, point out the laughably wrong ‘model’ that the White House is now using:
https://www.vox.com/2020/5/8/21250641/kevin-hassett-cubic-model-smoothing
They’ve since walked back that it’s a ‘model’, since they literally just used an Excel smoothing function with zero predictive capability, but good god I’ll take ‘uncertainty’ over ‘certainly wrong’ any day of the week.
You just could not resist the OrangeManBad urge, could you? Get some help before it takes over your life.
Oh, too late.
Not sure how pointing out that some guy who is an advisor to Trump used an utterly stupid Excel function with zero predictive power as a model is now ‘Orange Man Bad’ territory. President Trump himself didn’t factor into this.
This would be like dismissing criticism of an Obama advisor’s bad decision as racist because ‘Black Man Bad’ or some other nonsense.
“black man bad” , historically, was a Democrat talking point,
for a couple centuries.
That quip really didn’t help your credibility, although you probably were oblivious to it. And it certainly shines light on the shameful racism of many democrats.
Try again?
A few points, LC:
1. On a given machine, for the same inputs (including initialization and/or seed values for any pseudorandom generator variables the model uses), round-off errors will be the same for each calculation. On a given machine, provided the model in question is (a) correctly programmed, and (b) has not been changed, the outputs using identical inputs will be identical. Ferguson’s model apparently doesn’t do that. (You may see different results on different cpu architectures, however, since different cpu architectures can and often do handle rounding differently. Lorentz’s discovery that the underlying math used to model weather is chaotic was based on exactly that difference.)
2. If a model has been compiled using different compiler flags, you now have a different version of the model – not the same model. Ditto if any change has been made to the model’s source code prior to a recompile using the same complier flags.
3. If a model hasn’t been verified, you have no idea whether it is producing “correct” (from a calculation point of view) results or not. That’s the whole purpose of verification testing – to see that the model works “as built” correctly.
4. If a model hasn’t been validated using known correct data (either data obtained from dedicated validation testing or using known-correct real-world data), then . . . well, caveat emptor. The whole purpose of validation testing is to ensure that the model’s results, over some range of inputs/outputs, reflect real-world reality with some degree of accuracy. Absent validation, the results a model produces are of unknown utility because they may not even remotely resemble reality.
5. The paper’s author is correct when he/she asserts that the conversion of single-threaded code to a multiple-cpu environment can be a very difficult task. Synchronization issues in such a case can be a killer.
I’ll disagree slightly – in computational models, the ‘model’ is the code of the model, not a particular build of that code. You don’t have a different ‘model’ when you build a code with GCC 10 vs. GCC 8, and yet, even without bugs, you can easily get different results. But it’s still considered the same model. This is precisely why trusting a particular build of a model to give you the one-and-only-truth is absolute heresy. Statistical runs, with different compilers, optimizations, platforms, etc., should all be done, or at least demonstrated to give no drift in results outside of expected noise. That, or embedded quantification of error ranges, but that’s considerably more difficult.
And yes, conversion from serial to parallel code, be it via threads or task-based parallelism, can be difficult. Yet, it’s done every day, all day, in countless universities and national labs around this country and the UK. It’s not ‘Prove Fermat’s Last Theorem’ difficult.. it’s ‘Hire an experienced parallel computing software engineer’ difficult. Which is to say, fairly routine If you care about the validity and efficiency of your model.
If a model is being distributed as source vice as a compiled executable, it should specify in detail the compiler flags to be used in compiling same. Some compiler flags (the GCC -ffast-math compiler flag is an example) can cause problems if used – e.g., they can cause the model to produce different results if they weren’t used during testing. (That complier flag apparently screws with some of the normal mathematical order-of-precedence during execution, so code designed to use normal mathematical order-of-precedence may no longer function correctly when complied with that particular flag set.)
If a model is being distributed without specifying the compiler flags used during testing, I would ask which were used during testing. And I’d also be suspect of just how well the model had been verified and validated.
” Statistical runs, with different compilers, optimizations, platforms, etc., should all be done,”
Why? We already know using different whatever will yield different results. Using all those different things just adds more and different errors. Introducing multiple exogenous (nice word, eh?) errors is pointless if you are trying to show the validity or accuracy of a model.
It’s also more expensive in time and money. There are just way too many permutations of hardware and software to make that practical.
I would need to read of an actual case or three where that is done before I believe it.
Hey! I was going to say that.
Really.
I enjoyed the article, it even made me laugh out loud (at Ferguson et al.) occasionally. The author demonstrates a pretty thorough knowledge of programming, at least in my semi-educated eyes.
” But in any floating-point calculation, …”
Yes, true. Try expressing .3 in binary as an example. Pretty much every math and eng. student takes a course called “numerical analysis” which begins to explain these built-in inaccuracies.
“Non-deterministic outputs. Due to bugs, the code can produce very different results given identical inputs.”
The key word here is “very”. After the intro. course “numerical analysis” there are courses and content in other courses covering this problem in agonizing detail. The goal is to minimize these different results, even in “stochastic” programs.
Two areas of study is “sensitivity analysis” and “conditioning”. The idea is to determine how a small change in inputs affects outputs, and minimize any differences.This is an area which has been under study for decades, and remains so. Dissertations on this subject (error) are still done.
While not a programmer myself, I am familiar with programming and the math. and computer sci. behind it (enough to get myself in trouble, anyway). Sue Denim’s review, in my opinion, rings true and shows the hilarious efforts of Ferguson et al. to create some credibility for themselves in areas they have no expertise in; epidemiology, computer science, mathematical modeling, and a few others.
The models change because the circumstances on the ground changes and the information on the virus increases.
Early estimates were high because they were based on governments doing nothing.
They also prudently did not add wishful thinking into the model “such as assuming the virus would die off in the summer.
Now we know it spread slower in warm weather it R0 has change. Social distancing and lockdowns also changes the R0. As does estimates of compliance.
We are going to exceed 100,000 deaths by the end of May, and likely close to 250000 by the end November. And that will only be the first year of the virus.
The second year.will be worse because lockdown compliance will be out the window and the virus will already be widespread so when exponential growth resumes it will start much farther along the curve.
Unless the roll out a vaccine we could see a million plus by the time this is over.
Also, the evidence suggests that Covid deaths are under reported by around 50%.
But it will be a year or two until we know that with any real certainty
It changed, because they kept having to deal with being found out to have grossly exaggerated the likely case as worst case.
Real error tends to be randomly above and below the curve. if the corrections are always one way, someone is fudging.
Folks panicked and wrecked the economy. Other folks exploited it, like they always do, to wreck the opposition.
You just keep honkin the Kazoo, Komrade! Your credibility was shit before all this, and it isn’t growing roses yet!
“The models change because the circumstances on the ground changes and the information on the virus increases.”
If the models need to change with every acquisition of new data or parameters, they were obviously bad and useless. That, I believe, is our argument.
Well, in spite of Lar’s comments, there’s a fair amount of evidence that Ferguson is not just wrong, he’s an incompetent blowhard. From an article on the Brit website Spiked Online, located here: https://tinyurl.com/yahog29z
“In 2005 he said up to 200million people could die from bird flu – the final global death toll between the years 2003 and 2009 was 292. In 2009, the UK government based its ‘reasonable worst-case scenario’ for the impact of swine flu in Britain on Ferguson’s models, saying around 65,000 people could die. In the end just 457 people died. In 2001 Imperial modelling on foot-and-mouth disease shaped government policy, which was to cull six million sheep, cattle and pigs. Later, an expert in veterinary epidemiology said that modelling was ‘seriously flawed’.”
So, maybe Lars needs to explain why we should have listened to this fool in the first place?
The same reason that we need to listen to Michael Mann and his “hockey stick”.
Targeting Ferguson is a straw man on the issue.
The Ferguson model was not the one used to determine US estimates. We have several competing models by research teams. States have also created academic teams to model their own states.
Trying to claim this issue hinges on Ferguson’s credibility is like trying to claim the accuracy of all economic impact models hinges on the credibility of Larry Kudlow.
Ferguson is so damn irrelevant to the credibility of US models it is not just a straw man to prop him up as the center of gravity of aUS models, it is damn near a non sequitur to do so.
Ex’s original article was about Ferguson’s model, so your decision to make it a strawman is sort of changing the subject. Ferguson was quoted extensively in the press, so pointing out flaws in his predictions is not off topic.
But… since you wanna talk about US modeling, one used for many decisions is the IHME model from Wash. State Univ. There’s been criticism about it also: https://tinyurl.com/ybz6ubtk
“A widely followed model for projecting Covid-19 deaths in the U.S. is producing results that have been bouncing up and down like an unpredictable fever,…”
““That the IHME model keeps changing is evidence of its lack of reliability as a predictive tool,” said epidemiologist Ruth Etzioni of the Fred Hutchinson Cancer Center, who has served on a search committee for IHME. “That it is being used for policy decisions and its results interpreted wrongly is a travesty unfolding before our eyes.””
Modeling unknown systems is an opportunity to be spectacularly wrong. Real world data needs to be looked at also.
That’s Commissar’s problem… he’s all wrapped up in “empiric data” and “models” and not looking at real world data.
He’s like a first year resident… doesn’t look at or touch the patient, but can figure everything that’s wrong with then based on “the data”.
Right on.
“Empiric data” and “models”, especially mathematical ones, sound cool! Don’t argue with the science!
Changing the subject is par for the course for Koalemos (AKA Commissar AKA Poodle AKA Seagull AKA Cthulhu), George V. He seems to do that whenever someone provides facts undercutting his argument.
Commissar sounds like a replay of Joe Biden’s virtual “rally” a few days ago….
Trump
Covid Deaths
No Masks
Dead People
Trump
Here’s some hard hitting facts I came across this morning… I’ll let you come to your own conclusions:
https://pjmedia.com/instapundit/371621/#respond
https://www.nationalreview.com/corner/bad-state-decisions-about-nursing-homes-are-heavily-driving-the-coronavirus-outbreak/
https://www.frontpagemag.com/fpm/2020/05/did-lockdown-model-cause-half-coronavirus-deaths-daniel-greenfield/#.XrXq058-ZcM.twitter
Gee, do you think maybe enclosed ventilation systems in nursing homes have anything to do with that, as it does with Legionella?
Must be pure coincidence, right? Right?
Also why colder months being flu season is pure coincidence; enclosed indoor ventilation systems.
Contrast NY’s “you must admit those with the Wuhan coronavirus” policy for nursing homes with that of Florida – where the state government now requires a negative test before a patient returns from the hospital to a nursing home. Further, even before that policy was implemented hospitals in FL did not send those who’ve tested positive back to nursing homes.
Florida’s nursing home population is 70% of NY’s. Florida has 10% as many deaths among its nursing home population than NY does.
https://nypost.com/2020/05/05/new-yorks-coronavirus-nursing-home-deaths-didnt-have-to-be-so-high/
We had a nursing home not too far from here that was able to isolate those with the virus. This was early on, before much of anything was known about the disease. At least they had the good sense to separate all the COVID patients away from other residents, on separate ventilation systems, etc. Still, they had a lot of deaths there. They managed to keep those returning from the hospital isolated in a different area of the building.
Sending those with a disease, any disease, into the general population of a nursing home is unconscionable. Around here, they even take precautions during flu season. Or most do.