New Stanford Institute To Target Bad Science

ananyo writes "John Ioannidis, the epidemiologist who published an infamous paper entitled 'Why most published research findings are false', has co-founded an institute dedicated to combating sloppy medical studies. The new institute is to focus on irreproducibility, waste in science and publication bias. The institute, called the Meta-Research Innovation Centre or METRICS, will, the Economist reports, 'create a "journal watch" to monitor scientific publishers' work and to shame laggards into better behaviour. And they will spread the message to policymakers, governments and other interested parties, in an effort to stop them making decisions on the basis of flaky studies. All this in the name of the centre's nerdishly valiant mission statement: "Identifying and minimising persistent threats to medical-research quality."'"

Comments

The BBB For Science

[Russ1642]

Sounds like a great idea, but in reality it'll end up being untrusted and reviled by scientists. Set yourself up as THE authority on judging anything and the people you're judging will hate you because of your biases, conflicts of interest, lack of oversight, lack of accountability, and poor dispute resolution.

Re:The BBB For Science

[lgw]

Depends how "meta" they are. If their careful and question peer review practices and point out common methodology pitfalls, they might do OK. Better still would be to simply do science: science that refutes bogus published results through failure to reproduce the experiment as described. While that's absolutely key for science to work, no one funds it.

Re:The BBB For Science

[wbtittle]

They are only acting as the authority to point out the problems. There are huge problems in epidemiology. The really useful data gathered by epidemiology is not the positive correlations, it is the non correlations. This presents a rather ugly problem. The data that people find interesting are the positive correlations. With the exception of 1 or 2 studies, these are pretty much worthless. The data that shows a link isn't there is what is really useful. This is the source of all the bad research.

If you look at epidemiological studies, you find lots of RRs, HRs, and ORs (Relative Risk, Hazard Ratio, Odds Ratio). The confounding factor that is ignored is the Survival Ratio. The ratio of the survivors of doing something to the survivors of not doing that something. This number is almost always 99.99...% One exception is lung cancer and smoking. The survival ratio there is 92%. 92% of people who smoke their whole lives do not get lung cancer. (some simplification here).

Can't Come Soon Enought

[theshowmecanuck]

This can't come soon enough.

Re:Flashback

[JustOK]

I tried to make a photocopy of it once.

nerdish? wtf.

[rogoshen1]

Why exactly is "Identifying and minimising persistent threats to medical-research quality." even remotely considered "nerdishly valiant"??? That is a pretty important aspect of medicine that gets overlooked all to often by the pharma funded medical testing establishment :(

Re:nerdish? wtf.

[i kan reed]

You're reading "nerdish" as a negative adjective. I do not believe it was intended as such.

Re:wheeewwww

[NotDrWho]

It needn't produce any backlash. Let's just establish from the get-go that anything that contradicts my political beliefs is bad science.

Don’t throw a wet blanket on science

[Theovon]

It’s wrong to publish fabricated or falsified results, and people who do that should be slammed. There are other situations where people are being neglegent or hoping you don’t catch their slight of hand. For instance, there are the innumerable parallel computing papers that use O(N^2) algorithms to show a speedup on a GPU or supercomputer where there exists a serial O(log N) algorithm that runs faster on a PC. (No joke.) All of those sorts of things should be actively retracted.

However, what we don’t want to do is discourage publication of preliminary results that MIGHT be wrong. Honest, legitimate work that gets superceded should not be subject to retraction, and a wrong theory published can often inspire others to do a better job. When a researcher can say, “That was our best hypothesis at the time, and this was the most accurately we could represent the data,” then it should stand as a legitimate publication. Relativity and quantum mechanics supercede Newtonian physics, but that doesn’t mean we should retract everything Newton said.

Now, most people reading this will say “duh!” Because that’s obvious. All I’m saying is that we need to be careful to not create an environment where publication of preliminary work is discouraged in any way or where honest mistakes can hurt the career of an honest researcher. That would put a damper on science in general. The bar for retraction should be very high and require solid evidence of intentional wrongdoing.

Re:Don’t throw a wet blanket on science

[gstoddart]

For instance, there are the innumerable parallel computing papers that use O(N^2) algorithms to show a speedup on a GPU or supercomputer where there exists a serial O(log N) algorithm that runs faster on a PC. (No joke.)

Except that while there might be some problems which have O(log N) solutions as well as O(N^2) solutions, there are still things which still only have O(N^2) solutions, correct?

So if you can learn how to solve a known O(N^2) problem better (even if there is a known O(log N) solution), what you learn is still applicable to to other O(N^2) problems for which there isn't a known O(log N) solution.

I'm not sure what you're describing is evidence of malfeasance, or that they're working on solving a class of solution, and not necessarily that specific problem.

To me it sounds more like they're probably aware of the O(log N) solution, but that's irrelevant because they're looking at how to use parallelism to address things which are O(N^2), because there's many many of those.

So much of math comes down to solving an equivalent problem you already know how to solve.

Maybe they're figuring out how to address a problem which is O(N^2) by one method, so that once they know how to solve it faster with parallelism, they can learn how to solve other problems which nobody has an O(log N) solution for.

It may not be all about solving that particular problem, but that class of problem. Because mostly it seems like we've never figured out how to do real parallelism except for things which are classed as 'embarassingly parallel' because it already lends itself to breaking it up -- like SETI@Home.

Re:Don’t throw a wet blanket on science

[AthanasiusKircher]

When a researcher can say, âoeThat was our best hypothesis at the time, and this was the most accurately we could represent the data,â then it should stand as a legitimate publication.

Unfortunately, in many cases when people say this, what they often mean is: "This was the best a posteriori hypothesis we could come up with after trying out dozens of random correlations in our data to find something that could appear to be significant, and this was the most accurately we could represent the data after trying a couple dozen statistical measures to find something to make our minor 'blip' look more interesting."

In other words, it may well be a statistical fluke, but, hey, it is a "legitimate" way to represent the data.

All Iâ(TM)m saying is that we need to be careful to not create an environment where publication of preliminary work is discouraged in any way or where honest mistakes can hurt the career of an honest researcher.

Uhh, "honest mistakes" arguably should hurt the career of a researcher. If I'm an engineer designing a bridge, and I screw up my calculations, and the bridge falls down, my career should suffer. If I'm a researcher and I make a significant mistake collecting data or analyzing it properly or whatever, my career should similarly suffer.

"Honest mistakes" should NOT be punished as strongly as "dishonest" ones, but if you accumulate more than a couple significant "honest mistakes" in print, maybe people should start judging you.

I think perhaps what you really mean here is NOT that a researcher makes an "honest mistake," but rather that a particular experiment or set of research wasn't able to or designed to find some minor flaw -- not because of incompetence, but because technology or knowledge or whatever hadn't yet progressed to the point where anyone would think to look at things that way.

Basically, if a researcher screws up by not taking into account something already known (like misuse of stats or bad data analysis or bad methodology), that's a mistake -- but a researcher can't be expected to take into account that information when it isn't actually part of standard methodologies or whatever yet.

That would put a damper on science in general. The bar for retraction should be very high and require solid evidence of intentional wrongdoing.

I agree. The bar for actual retraction should be very high. But in this electronic age, the bar for significant corrections or qualifications which could be appended to an existing electronic document for actual "honest mistakes" should be lower -- it should be standard practice.

And, frankly, even if the research isn't mistaken, but is later superseded by more advances, we should start thinking about how to attach references to those sorts of things too -- lawyers do it when drafting a statute that replaces a previous one, to avoid confusion. Scientists should figure out a mechanism to do the same.

Re:Don’t throw a wet blanket on science

[gstoddart]

So is O(N^2) > O(log N) well that depends :)

Well, yeah, since it means on average this algorithm performs with around this cost.

Most of the time yes you can safely chose a log N and get better results. But not in all cases.

Yes ... but there aren't O(log N) solutions to all problems available to you.

What I was saying is maybe what they're studying is how to solve a known O(N^2) algorithm using parallelism to better learn how to tackle other O(N^2) problems. And that even if there exists an O(log N) solution for that particular problem, they may not be specifically tackling that specific problem, but how to work on the class of problem when you don't have an O(log N) solution available to you.

Even though you know someone has an O(log N) solution to a specific problem, you may be more interested in the generalities of how you solve O(N^2) problems, and the particular problem isn't really the point.

In other words, you're not doing the research to improve on that problem which already has an O(log N) solution, but to learn how to solve O(N^2) problems in general.

Re:Don’t throw a wet blanket on science

[johnwallace123]

And, frankly, even if the research isn't mistaken, but is later superseded by more advances, we should start thinking about how to attach references to those sorts of things too -- lawyers do it when drafting a statute that replaces a previous one, to avoid confusion. Scientists should figure out a mechanism to do the same.

If only there was a mechanism to refer to or cite previous work. I know... we can call them references, or citations! Awesome, I should publish a new paper telling everyone that they should use this system!!

This is where the money is short sighted.

[Anonymous Coward]

The Climatologists have the data - Compelling data.

And yet global warming has turned into this politically charged "issue" that has been created that way by moneyed interests who will not make as much money if certain policies to mitigate GW are implements - they won't lose, just not make as much profit.

What those people don't get, as things get worse - and they will - their interests are now in jeopardy. They will be labeled as the profiteers who paid for propaganda to slow down solutions. They will be labeled as folks who helped keep our heads in the sand and kept this needless "debate" going. Their money will be taken - lawsuits, fines, loss of business because they are liars.

I have one two words for them "Cigarette Industry".

They fought tooth and nail to hide, obfuscate, deny, gloss over, etc ... the truth. And in the end, they REALLY got it in the ass because of their actions.

If they just said up front, "Yeah cigarette smoking will kill you - one way or another - but it's out business and we're supplying what the market wants. And we are more than willing to switch businesses in order to save people and honor our fiduciary duty to our stockholders." they would be in a much better position now.

But they chose to lie and spread propaganda.

I think all of the folks who back anti-global warming propaganda should keep that in mind.

And let's just say that the one in a billion chance that global warming is just one big cock up of the scientific community (The odds are better that I'll win PowerBall 3 times in a row), we'll have cleaner air, water, less dependency on the whims of the international oil market, and our lives will be better - because we choose greener and cleaner energy.

Going with the Global Warming crowd is a win-win from my perspective.

Re:This is where the money is short sighted.

[wealthychef]

this has nothing particular to do with climate change, or any specific issue. Politicization is missing the point. The problem is bad science. We scientists need to clean our house before complaining about politics.

Re:This is where the money is short sighted.

[riverat1]

Is that your idea of science? "My cause is the right one, therefore it shouldn't ever be challenged."

No, if you want to seriously challenge climate science orthodoxy (or any other scientific orthodoxy) you need to put in the work first to really understand the science so you can intelligently challenge it in a scientific manner. Repeating past challenges that have been refuted many times already or not paying enough attention to what the orthodoxy is actually saying so you can address it directly just doesn't cut it.

For example the recent claims of 16 years of no warming. If you analyze the temperature records since 1998 statistically it's impossible to say whether the previous trend has continued unabated or if the trend is 0 increase in temperature. The period is just to short. But that doesn't stop climate science deniers from proclaiming it as evidence for the failure of climate science.

Same thing with the claim that climate models failed to predict the current pseudo-pause. If you understand how climate models work and how the results are presented as an average of many individual model runs you would know that they wouldn't be expected to predict such a short term deviation from the average.

So if you really want to challenge current climate science do the work, understand what the current orthodoxy is and come up with something that does a better job of explaining the evolution of climate. Otherwise it's just a bunch of hot air.

Re:Flashback

[HornWumpus]

http://www.jir.com/

Appears to be the same dude.

Re:Flashback

[sribe]

Does anyone else remember "The journal of irreproducible results."?

The Inheritance Pattern of Death

Infectious Diseases in Bricks

Behavioral Genetics of the Sidehill Gouger

Golf and the Poo Muscle

Oh, in answer to your question: yes.

Maybe I'll "Remote View" this

[Anonymous Coward]

Brought to you by the same people who gave us Putoff, Targ, and Uri Geller.

Re:Pretty much useless and wasteful

[TheCarp]

Are you familiar with the original paper?

> Can they target anyone with power that control unlimited resources, yet fabricate data on a
> daily basis

Maybe I am wrong but that doesn't seem to be what this is about. It is less about fabrication of data than it is about poor study design that lead to false results. From the original abstract:

Simulations show that for most study designs and settings, it is more likely for a research claim to be false than true. Moreover, for many current scientific fields, claimed research findings may often be simply accurate measures of the prevailing bias

This isn't about people in power with control of resources or about data fabrication. It is about making avoiding errors that lead to genuine reports of bogus findings.

Instead of this...

[Dcnjoe60]

The real problem isn't with shoddy research and researchers, the world has always had those. The real problem is the integrity of the journals that publish research. If they don't practice due diligence and report faulty studies, then they, the journals are at fault. The proper solution to faulty journals is to publish journals that have integrity and exercise due diligence. In a publish or perish world, not publishing shoddy research corrects the problem. What is needed is not the Stanford science police, but journals, symposiums, etc. with integrity that only allow the publishing/presentation of research that has been reviewed and vetted.

According to the latest study....

[HighOrbit]

This link is blatant right-wing propaganda, but funny as hell. Especially the one about fish.

http://www.consumerfreedom.com...

But on a serious note, todays NY Times had an "according to the latest study" acticle about a study that claims that all that stuff we've been told for decades about dietary fat being unhealthly is untrue. http://well.blogs.nytimes.com/.... Now since this contradicts several decades of observation, I tend to take "latest study" science with a grain of salt and give more credence to well verified (i.e. long term) science.

The problem with bad science is that it gets reproduced in the popular press (and popular imagination) even if it is later proven false. Case in point: the notorious vacination-autism fiasco. Another example is the "neutrino faster than light" results released a few years back in Italy. As Mark Twain said, "A lie can travel halfway around the world while the truth is still putting on its shoes."

You can never fully discount the possibility that the guy releasing the results of the latest study is an attention-whore looking to drum up sensationalism to have his 15 minutes of fame. Scientiest are human and subject to the same vanities as everyone else.

Bottom line, never trust preliminary results.

Science doesn't work?

[TapeCutter]

Should that be exempt from criticism?

Of course not, however you need more than just vague accusations, how about some actual evidence? Who are these greedy scientists and why do the criticisms sound like a creationist conspiracy theory? Who is paying for this "propaganda", what personal benefit do they gain from convincing people AGW is real? Why are these particular criticisms only raised on particular subjects such as AGW, evolution, and lung cancer? How is it that other scientists such as people hunting exo-planets are never accused of inventing planets "for the grant money"? Could it be because the findings from some branches of science threaten the power and purse of the rich and careless?

Is that your idea of science? "My cause is the right one, therefore it shouldn't ever be challenged."

The "cause" of science it to seek truth knowing you will never attain it. The "cause" of the billionaire neo-luddites is to make sure that critical thinking doesn't catch on with the general public.

securing grants, currying favor from academic mentors, generating press, enlisting public support, and so on.?

What exactly is wrong with any of that, does it not just add up to an ambitious scientist? Is the ambition of seeking the truth a bad thing in your eyes, or do you only see tax dollars going in one end and a "rich scientist" (lol) saying something you don't like coming out the other end?

What about all the pro-global warming propaganda

The pseudo-skeptic's reverse charge of propaganda from scientists is pure nonsense, sensationalism and exaggeration in the press is not "propaganda". Look at the technological wonder of the modern world around you for god's sake, propaganda is more than a mere lie, it a powerful psychological tool that convinces you that despite the futuristic world you find yourself in - (some) Science doesn't work.

Re:And the coughing is twice as bad!

[theshowmecanuck]

It's the dying that really gets you. One of my best friends died of complications due to measles when I was around 12.