Monthly Archives: June 2013

Supreme Politics: Scalia, Perry, and Windsor

Valid same sex marriages in the US will now be recognized by the federal government (so, Social Security benefits, joint tax returns, etc.) according to the Supreme Court.  There are interesting politics in the decisions today, though.

First, Chief Justice Roberts dissents, pointing out firmly that nothing in the DOMA decision (United States vs. Windsor) means that same-sex marriage is a constitutional right, or that states can’t keep the “one man and one woman” definition.  He also tipped the court’s hand in Perry, pointing out that they would dismiss it on standing grounds later in the morning.  
It’s Scalia, though, that has the really fascinating opinion here, one that I sort of agree with, in a weird way.  He firmly believes that the majority in Perry (the “liberals” plus conservative, gay-rights-friendly Kennedy) are planting this opinion as a step towards finding a constitutional right to same-sex marriage.  He spends a lot of time on this.
You see, Scalia would have dismissed the whole case on standing grounds, arguing that the US didn’t really want to win the case (that is, successfully defend the Defense of Marriage Act) and so there was no “case or controversy” — both sides agreed, so there is nothing to decide.
He accuses the US, and the majority in Windsor, of ginning up a fake controversy so that the the Supreme Court could write an opinion, binding on the lower courts, that would allow gay couples everywhere to challenge bans on same-sex marriage, even though that question was not before the court.  
You see, back when Kennedy penned the decision in Lawrence v. Texas (which declared so-called “anti-sodomy laws” unconstitutional) Scalia complained in his dissent that the decision would ultimately require that states recognize gay marriage (actually, his list famously went on to include prostitution, bigamy, bestiality and (*gasp*) masturbation, among other things).   Kennedy’s decision in Lawrence actually claimed it did not require that, but Scalia’s warning was quite prescient.  
Indeed, Scalia calls Kennedy out in Windsor for claiming it doesn’t mean same-sex marriage bans are unconstitutional, saying “I have heard such… disclaimers before” and citing Lawrence. This time he is even more explicit:

“In my opinion, however, the view that this Court will take of state prohibition of same-sex marriage is indicated beyond mistaking by today’s opinion.”

Whoa!  I believe Scalia.  He spends pages and pages backing this opinion up.  Gay marriage would be found to be a constitutional right if the question were to reach the Supreme Court.  Blow my mind.

But wait, that’s exactly the question the court decided today in Hollingsworth v. Perry!  But here there is a catch:  the standing issue came up again, since here the State of California, which had passed Prop 8, banning same-sex marriage, declined to defend the law in court, and appointed the proposition’s defenders to represent it in court.  In this case, though, 5 justices said that there was actually no controversy, since California wasn’t actually present in the case.  This is confusing: in Windsor the US showed up as a petitioner but agreed with the respondent, and the Supreme Court found 5 votes to say that’s OK, we’ll let someone else defend the law, just so long as your name is on the case.  But in this case because California wasn’t a party to the case and sent someone else instead, five justices said that’s not OK, and threw the case out.  
Standing is a fuzzy issue that lets judges dodge tough cases when they want to; this reeks of a dodge, a plausible but unprincipled stand to play a longer game on the subject.
So what happened here?  My bet:
Scalia and Roberts were sure that 5 justices wanted to make same-sex marriage legal everywhere.  Roberts, seeing this dynamic, got the tossed the case on standing grounds.  The weird thing is the five votes he got included Kagan, Breyer, and Ginsburg, who presumably were on the other side of the issue.
The dissent in Perry was written by Kennedy, who said not a word about same-sex marriage but blasted the standing argument, saying the court should have decided the case.  This is weird, because at oral argument he was very uncomfortable with the case, and wondered aloud if the court should toss it as “improvidently granted” (another cop-out option).  There was some speculation the Scalia had maneuvered to get the case heard at all exactly because he knew Kennedy would not want to decide the issue just yet.  
Further, Kennedy has Thomas, Alito, and Sotomayor on his side.  It is safe to say that Thomas and Alito would have found against same-sex marriage, and if we believe Scalia then Sotomayor is for making it national.  
So my guess is that Roberts and Scalia voted to toss the case on standing lest same-sex marriage become a national right.  Kagan, Breyer, and Ginsburg were perhaps happy to wait for a later date, perhaps when the court was more liberal, and especially since the decision in Windsor basically guarantees that result, anyway.  And perhaps they really bought the standing argument as an important principle, or perhaps they disagreed with Scalia and Robert’s assessment of the outcome if it had been decided.
So why did Thomas and Alito sign on with Kennedy?  Presumably because there was no downside, and they bought Kennedy’s argument.  Thomas is usually quite principled, and perhaps the standing issue was just really important to him.
Complicated! Bottom line on same-sex marriage:  it’s a federal, constitutional right in California, but nowhere else (huh?).  Scalia predicts that if the Supreme Court gets to decide the case, the court will extend the right to all of the US.
So, who’s the first same-sex couple to follow Scalia’s advice and file for marriage in a state that bans it, citing Scalia’s dissent in Windsor as evidence that they should be able to get married?

MARVELS-1: A case study in healthy paranoia in science (Part IV)

The saga of MARVELS-1 continues.


In part 1, I described how I was excited to discover that residuals to a one-companion fit were consistent with an inner planet orbiting the star MARVELS-1 in a 3:1 commensurability with the known brown dwarf, MARVELS-1 b.

In part 2, I explained our attempt to characterize the system, and our general unease with the system and our decision to get some Keck data to settle the issue.


In part 3, I described the universe of false positives we searched, trying to explain the data in some other way, and the reasons we were able to discount all of them.   I described how, in Jackson Hole, after conversations with Andrew Howard and Suvrath Mahadevan I finally built a toy model of contamination from another spectrum.  Here is what I found:

When we measure precise radial velocities, what we are doing is seeing how the wavelengths of various spectral features compare to their wavelengths measured at another epoch.  This tells us the change in velocity of the star, from the Doppler effect, from which we can infer it has orbital companions.  We measure this shift by building a model of the star (usually a “template” spectrum taken at another time) and asking “by how much must I shift this template spectrum to match the observations?  The shift that produces the best match gives us the measured velocity of the star.

This process can be thought of like a cross-correlation:  you “slide” the template spectrum (the model) across the “real” spectrum (the data) and multiply them at each position (shift).  The shift that maximizes the product of the data and model is the peak of the cross correlation function and your best fit velocity.

Now imagine that you have spectral contamination.  There is another, weak spectrum in present in your data at some other velocity, and when you slide your template across the data it will “catch” weakly at that velocity, giving you a weak peak in the cross correlation function.  But the brighter spectrum gives you a bigger peak, so it doesn’t matter.

Now imagine that the two spectra are almost on top of each other (at the same velocity shift).  The weak, secondary peak will create a “hump” on the side of the primary peak, and shift the maximum of the primary peak towards the velocity of the secondary one.  Suvrath calls this “peak pulling”.  There is also probably a similar secondary spectrum in your template, so really there are four CCF peaks competing!  But let’s focus on these two here.

I built a model that fit the data assuming that there was a secondary spectrum that “pulls” the velocities we measure towards its velocity.  This effect gets stronger the further apart the velocities are (there is no anomaly when they are perfectly aligned), but at some point the peaks become so separated that they don’t interfere with each other, so I had this peak-pulling effect weaken with distance as well (according to a Gaussian).  I let the strength of the effect, the width of the weakening Gaussian, and the velocity of the contaminating spectrum float as free parameters.

This model fit the data on the first try, almost perfectly.  I was stunned.

How on Earth could this produce a 3:1 resonance?  The model parameters have to be tuned just so.

Here are the original velocities:
Screen Shot 2013-06-20 at 10.07.47 AM.png

And here are the perturbations in my model:

Screen Shot 2013-06-20 at 9.50.25 AM.png

Here you can see the effect of the peak pulling.  At zero velocity, the spectra are aligned, so there is no perturbation.  As the orbital phase of MARVELS-1 b increases (the RVs from its orbit go positive) the contaminating spectrum “pulls back” harder and harder and the perturbative velocities go more and more negative.  Eventually, the lines start to separate in velocity space (at around an RV of 2 km/s, or phase 0.15 in the top figure), and the effect begins to flatten and even reverse.  The black points here are the HET RVs after subtracting my model’s best fit for the TRUE RV curve.
Below in a blue dashed line is the TRUE RV curve of MARVELS-1:

Screen Shot 2013-06-20 at 9.53.15 AM.png

The black line is the perturbed RV, that is, what we measured at the telescope because of peak pulling.

The red line is the best fit sinusoid to the measured velocities.  Because the measured signal is not a sinusoid, the best fit “splits the difference”, running high at first, but then missing low at the peak.  Count the number of times the red and black lines cross:  once at zero, once on the way up to the peak, then twice more on the way back to zero, then two more times on the other side.  Six crossings per orbit.  The difference between these two curves (what we would measure as “residuals” to a one companion fit to MARVELS-1) looks like this:
Screen Shot 2013-06-20 at 9.57.28 AM.png
and the actual residuals look like this:
Screen Shot 2013-06-20 at 9.58.42 AM.png
Bingo.  That’s how you get a fake 3:1 resonance from spectral contamination.
Next up: what could the source of contamination be?  As a sneak preview, the morals of this story will be: be suspicious of your own results, trust good students, and take your referee’s reports seriously.

Bechdel, Finkbeiner, Work-Life Balance, and sexism in Astronomy 101

There’s been some increased interest on the internet lately about the Bechdel test for women appearing in film.  The test comes from a comic strip in which a character explains that she only watches films that contain:

  1. At least two female characters
  2. That talk to each other
  3. About something other than a man

Depressingly very few films pass this test, which is partly a testament to the lack of female protagonists in film (few supporting characters in film have any role except to talk about the protagonist) but mostly a statement about how few well-developed female characters there are in film at all.

This is a good example of unconscious bias — most of this is probably not intentional on the part of screenwriters or producers selecting scripts (oops!  a second female character!  better change the gender there…), but an internalization of the market forces that create structural obstacles to films that would pass the test (the huge market for action films, the supposed difficulty of marketing a film with a female lead, the fact that the genres that have female leads generally have her talk only to or about men, other tropes and structures of successful films, etc.)  If you challenged them to write and produce films that satisfy the rule, I’m sure they could do it (couldn’t Pepper Potts and the Black Widow have chatted about the nature of S.H.I.E.L.D. in some plot-advancing exposition without damaging the narrative flow of The Avengers?)
Anyway, the point of the rule is not that a film is not sexist if it passes (my hypothetical Potts-Black Widow conversation wouldn’t have solved any problems) but to point out the ludicrously low bar that one can set and still trip up most Hollywood films.  
A similar idea is the Finkbeiner test.  Now, when I was in graduate school, the Finkbeiner Test was whether Doug Finkbeiner thought that your IDL code was sufficiently concise.  But I first heard of the sexism version after reading about the backlash to a New York Times obituary about celebrated rocket engineer Yvonne Brill (as someone, I think it was Josh Peek, pointed out, as a scientist, I would never get into a rocket built by a scientist.  I want a rocket engineer!)  Anyway, the obit opened with:
She made a mean beef stroganoff, followed her husband from job to job and took eight years off from work to raise three children. “The world’s best mom,” her son Matthew said.
Yvonne Brill, who died on Wednesday at 88 in Princeton, N.J., in the early 1970s invented a propulsion system to help keep communications satellites from slipping out of their orbits.

The best retort I found, exposing exactly what is wrong with this, was in a mock obituary of a Family Man Who Invented Relativity.

The Finkbeiner test applies more generally to science writing, and exposes the trope of writing about female scientists in the context of the difficulties of being a scientist and a mother/role model/human being at the same time.  The test is that the article must not explicitly mention that she is “the first woman to…” or mention the scientist’s: gender, child care arrangements, husband’s job, nurturing, being role model, or surprise at her field’s competitive nature.
The point of the Finkbeiner test is not that writing about such things or talking about them is necessarily sexist.  Indeed, there is a strong argument that we should be talking more about work-life balance, and especially work-family balance issues in the profession.  Talking about it in the popular press helps humanize scientists, which is also a good thing.
No, the point of the Finkbeiner test is to point out that such discussions almost never happen in popular profiles of male scientists, which exposes a double standard that is sexist, even if it is the product of unconscious bias.
Anyway, I started to wonder if my introductory astronomy class would pass the Finkbeiner test.  I do a “great scientists” theme and test on it, so I looked through my course notes to see which women I chose to include in “testable” material and what I wrote about them on the slides.  The women’s names appearing on my comprehensive final exam are:
  1. Marie Skłodowska Curie
  2. Henrietta Swan Leavitt
  3. Jocelyn Bell Burnell
  4. Vera Rubin
  5. Andrea Ghez
(I also had Jill Tarter on the test because she came to Penn State to give a public lecture last year, but she’s not part of my standard curriculum.)

Marie_Curie_c1920.pngAll of my descriptions passed with one exception:  I mention Marie Curie’s husband and daughter, and mention that she was the first woman to win a Nobel Prize.

I think, though, that in the context of the class these are forgivable.  I mention Pierre and Irène because they, too, won Nobel Prizes, and my point is to give some human flavor (which I do for all of my “great scientists”).  Besides, this was a scientifically important family, and that is cultural knowledge that I want my students to have.  
As for her being the first woman to win the Nobel Prize, that, too, is culturally relevant, and I only bring it up to follow it up with an ever more emphatic point, which is the one I test them on: she was the first person to win a second Nobel Prize.  I think this counteracts any semblance of a lower standard that the construction “the first woman to…” implies.
Anyway, I hope that I would have also mentioned similarly Nobel-winning family members of male scientists in my lectures.  
I also mention that Jocelyn Bell Burnell did not win the Nobel, and say that I think it was unjust.  I don’t put it in the context of sexism, though that is implicit.  But I make a point of all of the Nobel Prizes awarded for the discoveries I discuss, so it’s unavoidable, and I don’t think it violates the spirit or letter of the Finkbeiner test.
I also notice that my mini-obituary of Lynn Margulis also fails the Finkbeiner test:  at the very end (it’s at the end!) I mention that she was briefly married to Carl Sagan.  I think that’s forgivable, too — it was a factoid I was actually unaware of until she died, and I think if their genders had been reversed, I would have done the same.
OK, enough omphaloskepsis.  Next time: back to astronomy!