Thursday Throughput for 3/28/2019
So the Thursday Throughput starts with two long essays and then some links. Bear with me.
[ThTh1] Last week, NASA announced that they were going to have the first all-woman spacewalk in history to do some maintenance on the space station. This week, they had to cancel that because it turns out they don’t have spacesuits ready in the right size for the women. The internet has been slow-rolling an outrage over this, crying sexism. Even Hillary Clinton tweeted about it.
Except … you have to read a bit into the article or read this Twitter thread to find out what’s really going on. This entire nontroversy is a result of two things colliding: one good and one bad. The bad part is that NASA’s spacesuits were built in 1978 and only 11 of them are still in service. None are small and only two are medium size. Anne McClain originally was set to use a large suit. But during her earlier spacewalk, she realized a medium fit her better. Since it was impossible to get both medium suits ready in time for the spacewalk, NASA did the sensible and safe thing: changed the assignments.
The good part is that NASA has been making a concerted effort to recruit more women into the astronaut corps. McClain and Koch’s class was half women. So what’s really happening here is that a half-female astronaut corps is trying to use suits designed for the mostly-male corps we had in the 1970s. That’s the collision.
Can we just “build more suits” as Hillary Clinton snarked? It’s not that easy. They are tremendously expensive to build. They are tremendously expensive to develop, with some estimates as high as $250 million. A decade ago, Congress allocated money to develop a new suit for the Constellation program, but the Constellation program was eventually cancelled. NASA currently plans to use the existing suits through 2024 (the planned lifetime of ISS) but is unclear about what they’ll do afterward.
Frankly, I’m disinclined to take seriously the complaints of politicians on this. They made this bed. Both parties have seen NASA as something to gut for their pet programs and neither has seen fit to allocate the funds that would be needed for an aggressive manned spaceflight program. Mike Pence gave a speech yesterday calling for a return to the moon in five years but no one, and I mean NO ONE, is taking that seriously. I’m sure NASA would love to have enough suits to solve this problem. But without money, it’s not happening.
[ThTh2] The headline of this article — “Scientists rise up against statistical significance” — is a bit misleading. What’s going on is that more scientists are coming out against what I call the Church of P. It has been a tendency of science — particular the social and biological sciences — to proclaim that anything that has less than 5% probability of being random chance (a P value of less than 0.05) is a huge discovery and anything with more than that is garbage (my own field, astrophysics, uses a much tighter definition of statistical significance).
But the difference between P of 0.05 and P of 0.051 is not that huge. It’s just a number. And the Church of P has led to things like “p-hacking” where scientists try to find anything that crosses that threshold and proclaim it to be a result. Some scientists have done junk studies to demonstrate how this can be done.
Bayesian statistics can improve upon this but that field is … complicated. And I really don’t want to start a holy war in the comments between the Bayesians and the Frequentists. Just read XKCD.
Let me illustrate the point these scientists are making with an example from political polling. Let’s say a Presidential poll shows Donald Trump at 50% and Oprah Winfrey at 48% with a margin of error of 2%. The press will sometimes say they are “statistically tied”. But they’re not. That poll result indicates that there is at least a 2/3 chance that Trump is winning. It doesn’t tell you that Oprah is effectively tied with him; it just tells you it wouldn’t be a huge shock if she won. We saw this play our in real life three years ago when some statisticians, drunk on state level polling, said Clinton was a 98% lock to win while Nate Silver, more attuned to the uncertainties (and biases) set her odds at 70%. And, as he has endlessly reminded people, if the 30% chance doesn’t happen 30% of the time, you’re not doing statistics right.
What this petition is about is not doing away with statistical significance but destroying the Church of P and taking a more nuanced approach to statistical significance rather than using hard cutoffs. I agree … mostly. The only caveat I would have is that there is a much bigger problem: the bias in what studies get published. A study that shows something increases your cancer risk 20% (give or take 30%) gets published while a study showing it decreases it 20% (give or take 30%) is not. So I’m with them … as long as they also advocate publishing negative and contrary results so that we can get a real handle on whether a study is significant or consistent with statistical noise.
[ThTh3] Does binge-watching Netflix raise your chance of colon cancer? Hey, it’s time to talk about statistical significance again! In a study of almost 90k women, 118 developed colon cancer before age 50. Breaking down the numbers very small, they found the impact is somewhere between a 28% reduction and a a 75% increase for 7-14 hours a week and a 7-167% increase for 14 hours or moer. No, you don’t throw those results away (bigger studies have shown a bit more significance). But … I wouldn’t base anything on such a tenuous and tiny result. This is a case where the I think the statistics significantly understate the uncertainty.
[ThTh4] Is corn syrup fueling tumor growth? Color me a bit dubious on this one. Corn syrup is one of the demons de jour and we’re looking for anything we can blame on it.
[ThTh5] Some stars are young, some are old and some are really really old.
[ThTh6] Everyone, this is II Pegasi. II Pegasi, this is everyone. II Pegasi is a double star and the primary has tons of sunspots and sends off flares so powerful (a hundred million times more powerful than a typical Solar flare) that we can detect them 130 light years away. This is because it has an intense and tortured magnetic field, which the PEPSI instrument mapped out.
Yes, that is its name.
— LBTObs (@LBTObs) March 19, 2019
[ThTh7] A private Israeli firm is going to land the Beresheet (Hebrew for Genesis) spacecraft on the moon. Here’s a video it shot of the sun rising over the Earth limb.
Sunrise Video from #Beresheet ☀️ From the #spacecraft's point of view. In the video, #earth can be seen hiding the #sun & then exiting the same shadow created by the Earth and the sun's exposure. This process creates a kind of sunrise image! #IsraelToTheMoon @ILAerospaceIAI pic.twitter.com/y6IR80oz73
— Israel To The Moon (@TeamSpaceIL) March 24, 2019
[ThTh8] This is now becoming doable science: studying the atmospheres of extrasolar planets. And they’re turning out to be even stranger than imagined.