Slashdot Mirror
Billionaire Technologist Accuses NASA Asteroid Mission of Bad Statistics (sciencemag.org)
Taco Cowboy quotes a report from Science Magazine: Nathan Myhrvold, ex-CTO of Microsoft, is accusing NASA of providing bad statistics on asteroid size. Mr. Myhrvold alleged that scientists using a prominent NASA space telescope have made fundamental mistakes in their assessment of the size of more than 157,000 asteroids they have observed. In a paper posted to the arXiv.org e-print repository on 22 May, Myhrvold takes aim at the Wide-field Infrared Survey Explorer (WISE), a space telescope launched in 2009, and a follow-on mission, NEOWISE, which together are responsible for the discovery of more asteroids than any other observatory. Yet Myhrvold says that the WISE and NEOWISE teams' papers are riddled with statistical missteps. "None of their results can be replicated," he tells ScienceInsider. "I found one irregularity after another" Myhrvold says the NASA teams have made mistakes, such as ignoring the margin of error introduced when extrapolating from a small sample size to an entire population. They also neglected to include Kirchhoff's law of thermal radiation in their thermal models of the asteroids. Based on his own models, Myhrvold says that errors in the asteroid diameters based on WISE data should be 30%. In some cases, the size errors rise to as large as 300%. "Asteroids are more variable than we thought they were," he says. He has submitted the paper to the journal Icarus for review. However, the WISE and NEOWISE teams are standing by their results, and say that Myhrvold's criticism should be dismissed. "For every mistake I found in his paper, if I got a bounty, I would be rich," says Ned Wright, the principal investigator for WISE at the University of California, Los Angeles. Wright says that WISE's data match very well with two other infrared telescopes, AKARI and IRAS. To find out how accurately those infrared data determine the size of an asteroid, scientists have to calibrate them with radar observations, other observations made when asteroids pass in front of distant stars, and observations made by spacecraft up close. When they do that, Wright says, WISE's size errors end up at roughly 15%. Wright says his team doesn't have Myhrvold's computer codes, "so we don't know why he's screwing up." But Wright archly noted that Myhrvold once worked at Microsoft, so "is responsible in part for a lot of bad software."
Isn't that this patent troll?
But Wright archly noted that Myhrvold once worked at Microsoft, so "is responsible in part for a lot of bad software."
That hurts.
Here's a link to the paper. Seriously, does this guy think the WISE team are a bunch of idiots? I'm personally not qualfied to judge the details of the physical arguments in Myhrvold's paper, but I would give it high probability that he's full of shit.
The worst kind of data problem is when the data appear reasonable enough to not trigger suspicions but are actually dead wrong. I've been there, where I had a result that looked reasonable enough that I presented it at a conference. I later discovered that the result was wrong enough to invalidate anything I had presented. The problem was a software package I was using, a programming and data analysis tool called NCL. Had a function worked as documented and in the examples on the website, my results would have been fine. But because of the bug in NCL that probably still hasn't been fixed, my results looked reasonable enough but we're wrong. Thankfully I discovered the problem before trying to publish the results.
NASA says their results look reasonable. However, if their methodology has errors, it might be a case of reasonable looking results that are wrong. I think it would be worthwhile to look into the concerns and see if any of them have merit.
The last statement in the summary is completely uncalled for. The ad hominem attack does nothing to defend NASA's methodology. It only serves to try to discredit the criticism. That's the biggest thing I have a problem with. If you're convinced the methodology is correct and that the concerns are unfounded, that's enough to fend off the criticism. That the critic was once an executive at Microsoft is totally irrelevant.
He has also attacked other researchers in other fields: http://mobile.nytimes.com/2013/12/17/science/earth/outsider-challenges-papers-on-growth-of-dinosaurs.html
Shame he's been unable to locate the stick lodged in his butt.
" we're witnessing a new class of rich assholes who think they're superman or something"
er, no, that class is as old as civilization. We used to call them kings or lords.
Pain is merely failure leaving the body
Do we find the science too complicated? Too busy to actually read the papers? Too lazy to do a little digging? Never learned how to do the math? Never mind. We can always pick a side and run down the character of any and all opponents. It's quick. It's easy. It's fun. Science be damned.
Oh, please.
None of us, even the most scientifically sophisticated, is capable of developing sufficient expertise in every field in order to personally judge the scientific merits of technical arguments in highly specialized fields. Maybe one or two such fields, if we work very hard on it. This is why we rely on the opinions of experts. Putting every random crackpot who advances an argument on the same footing as established scientists in the field is false equivalence. Yes, every once in a great while, an outsider can point out an error being made by subject experts. But, 99.999% of the time, they're full of shit. The burden of proof here is on Myhrvold.
IIRC HPFS failed because it was part of OS/2, and because OS/2's file system requirements - that HPFS met - were considerably simpler than NT's.
IBM eventually dropped it in favor of JFS, but that was because it was owned by Microsoft, and (like all file systems) had hard limits on file and partition sizes that needed to be resolved by the mid-1990s. Had Microsoft and IBM continued to co-develop OS/2, it's quite probable HPFS would have continued to evolve too.
I don't disagree with the rest, but I'm not seeing a good reason to knock HPFS. It was the OS/2-NT politics that killed it, technically it was exactly what it said it was, a (reliable, stable) high performance file system.
You are not alone. This is not normal. None of this is normal.
It reminds of the time Tom Cruise told Matt Lauer that "There is no such thing as a chemical imbalance in the body.", and "You don't know the history of psychiatry, I do.".
The overpaid and entitled, with a pulpit to stand on and a microphone in front of them, will always feel the need to speak from a position of equal parts confidence and ignorance.
I'm sorry, but your opinion seems to be wrong.
The burden of proof here is on both sides.
Really. What the fuck does that even mean?
WISE has publicly released its data, and published multiple analysis papers in peer-reviewed journals . What, exactly, more do you expect them to do?
He has not published his alogorithm source code or calculation so the same apply to that guy. On the other hand result in agreement with other telescope and experiment give some more surety. So we have 3 results similar and 1 guy which makes some calculation and tell us they are all 3 wrong in effect and riddled of error " Yet Myhrvold says that the WISE and NEOWISE teams' papers are riddled with statistical missteps" "fundemental mistake" so excuse me if I pardon the "he made bad software".
C. Sagan : A demon haunted world:
http://www.amazon.com/gp/product/0345409469/
visit randi.org
That's kind of an interesting take on it. I wasn't sure what his motivation was; it read to me like Myhrvold was trying to say there aren't as many asteroids in the regions in question as NASA was trying to claim.
Is that what you picked up from his rant? That he's trying to say NASA is underestimating the size of the objects?
(Deciding to read TFA & his actual paper, now, to see what's going on...)
[...]
Well, it looks kind of odd to me. In the article, he says he was approached by B612 but stresses that he did not give them any money. He says "they came looking for me and my millionaire friends". Okay, so is it safe to assume there's a possibility that Myhrvold has taken a personal interest in B612's proposed model and is out to shoot down their competition?
In his paper, he mentions Sentinel (B612's project) in a terse paragraph but just before that he mentions LSST in a sort of verbose fashion. Then he goes on to write even less about Sentinel, and there he mentions B612's search for private funding. So maybe Sentinel isn't his pet project -- maybe B612 pissed him off during fund-raising (or maybe for soliciting funds from him at all.) Which would make sense if he'd already familiarized himself so much with LSST and had already adopted it as his pet project.
He cites the National Research Council as having determined that "LSST offered the most cost-effective and lowest risk approach". He seems concerned that LSST won't be finished on time, and therefore now I'm guessing that he hopes he can prove the other projects wrong so that LSST's long shot has a better chance.
(Sorry, writing this as I read his paper)
A few paragraphs later, he complains that simulation code used by the various projects isn't available for public scrutiny. I have to side with him on that much. If it's publicly funded, maybe at some point there should be a fundamental basis of experimentation that's also publicly available. It's sort of disheartening to read that each individual project is working from potentially grossly different simulation models. Hasn't some academic body somewhere already come up with the best model for these projects to use? Shouldn't that have been the first goal of the NEO search community?
He also mentions that each project can also add code simulating the results that competing projects might come up with. That's interesting, too. It sounds like the entire thing is very highly political. How many teams are publishing simulation results that downplay the accuracy of other teams? That doesn't seem very academically sound, at all.
He then goes on to say what I just concluded (that it's not very academic) and says exactly what I was also thinking:
"Ideally, the community would produce an open model that can simulate the NEO search performance of IR and visible-light telescopes, whether based on the ground or in space, with consistent assumptions and consistent input distributions of NEOs."
My sentiments, exactly, and I'm still just going paragraph by paragraph, here.
Later on in "Asteroids In Reflected Light", he selects his favorite functions, does some integration, and then iterates that theoretically derived functions that might be in use haven't been applied to a sufficiently large set of data. He says plainly that he prefers to use an older standardized model that is no longer the de-facto standard because the newer standards also suffer from lack of experimental data.
So, wait a minute. He has all this time, why isn't he simply trying to get more telescopes to focus on supplying the experimental data needed to make the newer standard (H, G1, G2) more immediately useful?
Well, he's just going on applying the assumption that the data from a known less accurate standard can be relied on within some margin of error to show whether newer models are accurate or not.
It seems a logical assumption that the margin of error that should be applied to the older model is a wide open variable. We're talking about tumbling rocks
"Stratigraphically the origin of agriculture and thermonuclear destruction will appear essentially simultaneous" -- Lee
Getting rich requires a certain skillset. There's little or no evidence that that skill set transfers over to other areas, like governance, or science for that matter. This glorification of the rich man as some sort of Nietzschean superman is absurd.
The world's burning. Moped Jesus spotted on I50. Details at 11.