Tuesday, December 20, 2011

You're killing me, NASA

The Kepler mission is doing pretty well. Recently they have discovered a couple of new planets. Fine. However, I'm pretty sure the only thing we actually know about these planets is that they are probably spherical. The artist conceptions take it a little too far. Why not go all the way?

I have a separate problem with the line-up. At first glance it looks like a bit more than a 3% difference in radius, but the south poles are all lined up for the other planets while Kepler-20f is bumped up a bit. I have no idea why they did this since I thought the similarity to the radius of the Earth was the interesting fact.

Saturday, August 6, 2011

Charity for the straw man

This person is a professor of something. That is not to say that I don't know what he is a professor of -- public affairs and economics in fact -- but that he passed the GRE at some point (likely with pretty good scores), graduated from grad school, probably got a post-doc or associate professorship, and became tenured. And now he starts a blog post with two regressions with R^2 = 0.21 and R^2 = 0.09 with n < 10 and then concludes that it justifies something besides inconclusiveness. That is our standard beef here, but not my biggest beef today. I have a meta-beef.

The other professor (Sumner) being criticized by this post only made the claim that deviations from Okun's law were small, hence unemployment was basically explained by NGDP. Therefore showing that deviations from Okun's law are within the error of the model as Chinn does proves nothing. That is the original point. You do not show someone's point is flawed by demonstrating their point.

I will be (uncharacteristically) charitable now to get to the heart of the meta-beef. It could be that Chinn is trying claim the model does not work better with revised data. I'll call this the uncharitable form of Sumner's claim. However, the original curve inside the error of the model; it is impossible to work better. Chinn is setting Sumner up to fail by using the uncharitable form, but writes with the pretense that the uncharitable form could be correct --  that the uncharitable from is decidable given the data, and therefore you can set out to demonstrate it in an unbiased, scientific manner.

Of course, writing with the pretense that uncharitable form is decidable you'd actually have to allow that it is possible the model works worse with the revised data. That means you would have to consider the charitable form of Sumner's claim: given the revised data the model still works just as well. Since GDP revisions tend to get smaller over time, Sumner can say now that the revisions are in, we still have the same picture. That while the GDP vs employment data could have gotten revised to deviate from the model, it did not; that is Sumner's "final nail in the coffin". No one will likely ever produce data that will show that Sumner was wrong.

I know, I know: uncharitable representation is an age-old device in science. But given an uncharitable representation of Simplicio's Aristotelian position, Galileo's arguments in favor of Salviati's Copernican position did not imply a charitable representation of the Aristotelian position because both sides did not believe in the same underlying model. The underlying model was the debate. Here, Chinn and Sumner are both using the same model. The conclusions are the debate.

Sunday, May 29, 2011

Fauci and McNeil: continuing adventures in bad science

WTF?!! [spit] AAAAAAAAAHHHHHH!!! [spit] #$%&!!!!!

What pharmaceutical company is paying to plant these articles?

This has the added outrage of not just being terribly unscientific but also morally suspect methodology. Not only is sqrt(1763/2) ~ 30, but they stopped the trial early when the difference started to appear significant. I can prove any coin is unfair if I stop flipping it when it starts to appear unfair.

Dr. Fauci and Mr. McNeil were apparently at it yet again ... we've caught them before:

Mr. McNeil also has this recent article. There is a theme involving the University of North Carolina, but they do have a center for AIDS research there. However, "Pharmaceutical Product Development, Inc." was a major donor to local races in North Carolina last year, though I haven't found any particular connection to any of the drugs in the trials or a ton of references to AIDS on their website. They have provided "research management" to NIH/NIAID (Dr. Fauci's organization) specifically for HIV/AIDS research. All of this could be plain "synergy" and the poor scientific methodology plain confirmation bias. I don't think there has to be something nefarious going on here.

I wonder if there isn't a deeper connection to this article. Pure speculation on my part at this point, but a couple of my favorite quotes ...
"We've never done ghostwriting, per se, as I'd define it", says John Romankiewicz, president of Scientific Therapeutics Information, the New Jersey firm that helped Merck promote Vioxx with a series of positive articles in medical journals. "We may have written a paper, but the people we work with have to have some input and approve it."

Underlining mine. And ...
Alastair Matheson is a British medical writer who has worked extensively for medical communication agencies. He dismisses the planners' claims to having reformed as "bullshit".

Spittle-flecked ire, indeed. Fauci and McNeil may be the worst collaboration of a scientist and a journalist since Levitt* and Dubner.

*I'm using scientist in loose sense here. Levitt is an economist, but a lot of economists use the scientific method. It is even looser because I'm pointing out flaws in the use of the scientific method.

We have another contender for the worst graph ever made

Or should I say three? I don't even think I need to say anything.

The reigning champ has the added benefit of using more complicated methods in the disservice of the pursuit of knowledge, but the new one has the added spice of moral outrage of poor scholarship in the service of the elite.

Saturday, February 5, 2011

Against Moore

So I was reading this, and then this, and then this, etc. Having spent most of my academic career (and now career) in male-dominated fields, I'm well aware of many of these gender issues. The problem is probably intractable until we reach a post-gender society where reproduction is either unnecessary or controlled by the state.

My issue is with philosophy. (Still.) And particularly Moore:
... it would be better for a beautiful planet to exist than an ugly one even if there were no one around to see it ...
 Let's replace beautiful/ugly and better/worse with positive/negative (gets us away from that whole gender morass). It would be more positive for a positive planet to exist than a negative one even if there were no one around to see it. The first half is a tautology; the second half states the tautology is true (duh) regardless of the existence of a choice function and hence independent of the axiom of choice. But definitions are independent of the axiom of choice. That I am unable to select a planet from a set does not mean given a planet and a function that assigns each planet onto the set {-1, +1} I can't find its image in that space.

It would be interesting if Moore considered the consequences of the axiom of choice, whether we can see it or not does have an effect in mathematics (and quantum mechanics, depending on who you talk to). But usually weird ones, like the Banach-Tarski paradox.

And this:
It's raining but I don't believe that it is raining
Under most circumstances, I believe this means the speaker is surprised it is raining. It is an emotional utterance; it is not weird or nonsensical in any way. People do not contain within them fully consistent sets of heuristics.  Supposedly Wittgenstein considered this Moore's greatest contribution to philosophy. This is another example of the fallacy of the preeminence of human thought that philosophy falls for so often.