galaxy concentration parameter

As much as a critic as I am of made-up statistics, Mykytyn and I spent part of the day measuring concentration parameters on our models of (angularly) large galaxies in the SDSS. We needed to do something like this because our galaxy models are not simple enough to have their radial profile described simply by any simple combination of fundamental parameters. We are working on fitting models to every galaxy—from any catalog—bigger than 2 arcminutes.

GALEX graphical model

I worked out a graphical model for the GALEX photon project, and it reminded me that in order to operate on the photons, we need a spacecraft attitude model, a focal-plane mapping model (WCS), a point-spread function model, a model of the focal-plane sensitivity map, and a background photon rate model. With all these in place we can do any time-domain projects. I also thought a tiny bit about how we might release these functions to the public when we release the data. In the meantime, Greenberg and Gertler are working on approximations, and re-discovering our transiting sources (which we still have to write up, Schiminovich!). In other news, David Mykytyn arrived in Heidelberg today for a week-long sprint on the Atlas.

Herschel emission predicts visible extinction

It's true, folks: The dust that radiates in the infrared is the same as the dust that absorbs in the visible and ultraviolet! Some may say this has been known for decades, but I successfully confirmed it today by comparing the dust models I can make for multi-band Herschel data on the M31 disk with extinction maps made from HST-imaging visible and UV stellar colors by Karl Gordon and Julianne Dalcanton of the PHAT Collaboration. My comparison is purely qualitative at present but I hope in the next week or two to make it quantitative. I want to find the combination of emission-inferred dust column, temperature, and emissivity parameter that best predicts the visible extinction; I then hope to find that it is only the column that is heavily involved in the prediction. That would be a great success of my (completely trivial, null) dust absorber and emitter model. I discussed all this today in one of the MPIA lounges with Ben Weiner (Arizona), Brent Groves (MPIA), Karin Sandstrom (MPIA), and Rix, with cameos by Thomas Henning (MPIA) and Tom Herbst (MPIA).

GALEX photons, odd quasar, etc

In MPIA Galaxy Coffee, I talked about our project to liberate all the photons GALEX has ever seen. I got lots of good comments and questions. After me, Decarli (MPIA) talked about one of the odder binary-quasar candidates that Tsalmantza and I found. The system is so odd that Rix commented that we should send it to Halton Arp: It looks like a galaxy ejecting a quasar or vice versa! As I have commented here before, our search for binary quasars has shown how rare they are: There really are no clear examples of two quasars orbiting each other at thousands (rather than hundreds) of km/s or faster. Maybe there shouldn't be such objects, but it is amazing that there really are none among the tens of thousands we have searched. In the afternoon, Weisz and I worked on some of his PDMF-fitting code.

2-d spectroscopy

In a low-research day, Ben Weiner (Arizona) showed me his impressive redshift survey using HST in objective-grism mode. The data are beautiful and the extracted spectra provide redshifts far fainter than anything I have ever seen.

a causal model for Milky Way disk stars

Bovy, Rix, Foreman-Mackey and I gathered in the "vdB lounge" of the MPIA to discuss a graphical model for stars in the Milky Way disk, in part to guide our thinking about projects and in part to make a possible illustration for a review that Bovy and Rix are writing. There was some argument about where to put the extinction, and what to consider the observables, with Bovy and Rix making the case that since they are given (by observers) temperature and gravity, these should be seen as observables, and me arguing that since these are coming from medium-resolution spectra, the spectra should be seen as the observables. This difference doesn't matter much to the inference, but part of my prejudice comes from the idea that graphical models can be thought of as causal models. Therefore I want my graphical model to represent the causal structure of the problem where possible.

Along similar lines I discussed with Beth Biller (MPIA) her projects to do exoplanet direct imaging campaign simulations based on data-driven models of the exoplanet population. Here one of the interesting things is that non-detections can be just as constraining as detections, and she has a nice framework for making all that happen. This is strongly related to exoplanet population modeling under consideration by Foreman-Mackey and myself.

modeling multi-resolution data sets

I worked today on our Herschel side project, in which Lang and I attempt to model the dust emission in the M31 disk at the resolution of the Herschel 100-micron images, but using information from all the bands out to 500 microns, which are much lower in resolution. It all seems to work! But we need ways to test the results. Ideas that have come up include: Check that we reproduce the extinction map. Check that we predict well 70-micron and 1-mm data. Check that we don't find strong A_V vs temperature variations. Any other ideas? We get good residuals internally to Herschel, but obviously external tests are much more valuable.

seminar

I gave my Hauskolloquium at MPIA today. I think I ended up convincing everyone to just coadd their damned data, even though my intention was just the opposite!

non-negative image priors

Foreman-Mackey and I sprinted on The Thresher in the hopes of giving me something to say in my Hauskolloquium here at MPIA tomorrow. We figured out something (already known) about non-negative priors: They can lead to severe biases. Consider, for example, PSF fitting with non-negative priors. If there is any noise in the system at the outkirts of the data used to find or fit the PSF, a system with non-negative priors can only capture the positive excursions; it can't also capture the negative excursions. This leads to two kinds of biases: One is that the PSF inferred with non-negative priors is always fatter than the PSF you would obtain by normal linear least squares. The other is that the PSF never really goes exactly to zero in the wings; it just can't. These problems arise, in my view, because we are doing point estimation; if we were carrying forward a full posterior PDF for the PSF we would be fine, but we just don't yet know how to do that! We came up with short-term hacks to deal with our problems and get me ready for my talk tomorrow, which is about extracting information from collections of images.

PCA, hst pixels

Long discussions today between Sandstrom (MPIA), Groves (MPIA), Kapala (MPIA), Weisz (UW) and me about inferring the total mean intensity from a patch of the sky from the pixel photon counts in a large patch of HST imaging. I think the problem is hard enough—given zodiacal, Galactic, and extragalactic backgrounds (well the first two I guess are foregrounds)—without the additional problem that we don't understand the HST pipeline processing. I got even more upset about the idea (encoded in DrizzlePac) that images are grids of buckets of photons; this view leads to ideas like "transformations that preserve flux" (when in fact transformations should preserve intensity), and makes all interpretation extremely sensitive to pixel solid angles, which may or may not be properly multiplied in (with an intensity image, failure to multiply in would be failure to calibrate; not so when you are just counting photons).

Collaborator Tsalmantza was shocked to hear me talking about PCA with the exoplanet direct imaging group (headed by Brandner) since I have devoted quite a bit of my last few years attacking PCA. I guess I now have to agree that PCA is sometimes useful. It is never Mr Right but it is sometimes Mr Right Now. Everyone in the direct imaging world is talking PCA right now, I think in part because we have all reached the limit of what means can teach us. In Brandner's group meeting we also talked about finding young M stars, which are great exoplanet search targets because they are not so luminous and when young the planets might still be hot. Young M stars have activity that give them line, UV, and X-ray emission; the planets might too; that seems like an interesting thing to think about!

writing, Thresher, PanSTARRS, HST

I didn't get much done today after a nearly-all-nighter helping with the Fergus paper on high-contrast imaging and various papers by Bovy and Bolton. However, I did pair-code a bit of The Thresher with Foreman-Mackey; we showed that we can take a small number of PanSTARRS cutouts with different seeing and make a coadded version that is higher in signal-to-noise and resolution than any of the original images from which it is built. (Phil Marshall made the image below, and also noted that we are "over-deconvolving" which is a problem I have to agree.) That could be useful to the survey team. Late in the day, Maria Kapala and I beat our heads against the HST data reduction pipelines used in the HST Archive and the PHAT HLSP Archive. There seem to be inconsistencies in the image pixel histograms that are very difficult to explain by any reasonable model of what these pipelines do.

rigid disk kinematics, habitability

After a morning of beating our heads against intransigent code, Foreman-Mackey and I spoke with Rix and Lisa Kaltenegger (MPIA) about their project to do a more responsible job of assigning probabilities of habitability to noisily observed tiny planets. The issues include: The observational uncertainties can be bad when projected into the quantities of interest (insolation and size). Some aspects of the problem are essentially unobserved (albedo and existence of an atmosphere and rockiness). Some planets have mass and radius observed, some have only one or the other. We discussed in general, and Foreman-Mackey is going to think about whether to take it on.

At lunch we discussed Bovy's excellent (not yet released) manuscript on the Milky Way disk using SDSS-III APOGEE, which is based on measurements of radial velocities all over the whole disk (APOGEE is an infrared spectroscopic survey, so it sees through a lot of the dust). Even marginalizing out distances to the stars, he can obtain a few-percent measurement of the rotation curve! One fundamental and obvious yet surprising point that came up is that in radial velocity the rotation of a perfectly cold, rigid disk is unobservable! Several of us laughed to realize we had never thought of that previously. Bovy's method therefore relies heavily on the asymmetric drift model—the fact that the lag of a population relative to the cold-population rotation curve is a function of velocity dispersion. But since he can see most of the disk, and since he is marginalizing out troublesome distance estimates (and also giant/dwarf classification), his results are the best ever. We discussed the comparison with future maser studies but I think that the much more precise maser measurements might not help that much: The masers and other young stars have a lag that is not easily understood in terms of the theory of asymmetric drift, as we showed here and here. Rix pointed out that it shouldn't be easily understood, because the masers aren't expected to be an angle-mixed population.

PHAT Camp, day five

I worked on wording, equations, and pseudo-code for completeness inclusion in the various PHAT projects underway. Gordon and Weisz successfully ran the PDMF fitting code from end-to-end on both simulated and real clusters. Also Maria Kapala (MPIA) came by and we looked at the pixel histogram in various PHAT UV HST images of the M31 disk. She is looking to see if it is possible to measure the light from the unresolved and undetected stars in the UV as a shift in the pixel histogram in the UV images. The histograms were nearly incomprehensible, making me wonder if MultiDrizzle is messing up the data. Of course I have never been fond of Drizzle-like algorithms, because they treat the data as overlapping square bins of photons and not samples of a pixel-convolved-psf-convolved intensity field. More on that here next week (for Brewer's sake if no-one else's). We may try to chase down the reasons for the pixel histogram issues next week.

PHAT Camp, day four; optical interferometry

Foreman-Mackey and I worked out the differences between, as it were, the Bovy view of the world and the Foreman-Mackey view of the world (well, the world of variable-rate Poisson processes subject to censoring). Of course the two views are mathematically identical in the end!

As work continued in PHAT land, I escaped for an hour to see an outrageous talk by Andy Skemer (Arizona) on high-resolution imaging with the LBT. He showed an absolutely beautiful interferometric image at four microns of a star made with the LMIRcam interferometer. It looked like what we see in textbooks! He also showed a video of the image over time, showing that the phase difference between the telescope's two mirrors varies stochastically with time, probably because of atmospheric variations. Foreman-Mackey and I discussed whether we could model these data—the individual-telescope AO PSF is so stable, the different images have only one degree of freedom, which is the unknown phase difference between mirrors. That's a one-parameter family of PSFs and 0.04 arcsec resolution in the good direction.

PHAT Camp, day three

I worked a bit more on the math behind variable-rate Poisson processes with censoring; the final tidying up on the work we have done over the last few days. In the meantime, Weisz and Gordon worked on getting code to infer the present-day mass function in various M31 stellar clusters. So all goes according to plan. In the afternoon, Christy Tremonti entertained Hennawi's group meeting with some very strange, very compact, very energetic post-starburst galaxies. They seem to have enormous ultraviolet fluxes but no emission lines. How is that possible?