Galaxy Zoo Starburst Talk

"021020" objects: some analysis of photometry errors

  • JeanTate by JeanTate

    I stumbled into this while trying to show that the "potentially problematic objects" (ppos) of the 'star contaminates photometry, spectroscopy' kind are objectively different from 'clean' objects, by looking at the data not the images. Among the 0.02 < z < 0.10 AND Z brighter than -20.0 objects (hence "021020"). Zooite mlpeck showed that most of such 'star ppos' within ~0.035' of the fiber center do, objectively, contaminate the spectra. But that ~none of the rest do. All this is in the Potentially Problematic Sources in 'Subset 2 -- 1149 Source Sample' thread.

    The data: the five "fiberMag" values - one each for the u, g, r, i, and z bands - and the five "cModelMag" values (ditto), and their PhotoObj errors (e.g. "fiberMagErr_g"). For the 1043 QS and 1090 QC PRIMARY objects in DR8*. I got these from the SDSS database by running some CasJob queries.

    In the next post I will plot the distributions of these ten errors. Then I will discuss the distributions, and what they may mean for the Quench project.

    *there are six QS and six QC objects in the Quench catalogs which are 'missing' in DR8; details in a post in the Potentially Problematic Sources in 'Subset 2 -- 1149 Source Sample' thread

    Posted

  • JeanTate by JeanTate in response to JeanTate's comment.

    enter image description here

    The scales are logarithmic, and where the Number in any bin is zero, I have replaced it with 0.1, so nothing 'goes off the chart'. The upper bound of each bin is 1.94 times its predecessor's; the upper bound of the first bin 0.01 mags. The vertical line is at a value of 0.3 mags, the threshold I used in my ppo analyses.

    The distribution of u-band fiberMag errors, for the 1143 QS objects, is very different from those of the other four bands.

    Here's the same plot, but for the 1090 QC objects:

    enter image description here

    Perhaps some minor differences, but fiberMag errors seem to have pretty much the same distributions; for this parameter, QS and QC objects seem to be ~the same.

    Time for cModelMagErr distributions; first QS:

    enter image description here

    Wow! 😮 Very different! For the g, r, and i bands the numbers are down to the onesies by ~0.05 mag, and the z-band isn't far behind. Even the u-band distribution seems to have shifted to the left.

    Is the QC cModelMagErr ~the same, as for the fiberMagErr distributions?

    enter image description here

    😮 😮 😮 Not going to need to do a KS test here!

    Time to remind ourselves what we're looking at.

    The QC objects were selected from all SDSS objects specifically so they match the QS ones in redshift and log_mass. You would think, intuitively, that there'd be no reason for the photometric error distributions to be different (OK, fiberMagErr distributions may be different from cModelMagErr ones, and each filter may have a different distribution).

    What is going on?!?!

    What is it about QS objects that leads to their cModelMagErr's being so small, compared with a sample of objects which have the same redshift and log_mass?

    It can't be errors in the SEDs their spectra record: you could swap the QS and QC fiberMagErrs and you'd never notice*.

    Anyone?

    *the whole point of fiberMag estimates is to do photometry on the light that goes down the fibers, to the spectrographs; similar fiberMagErr distributions imply very little in the way of any differential systematic effect here

    Posted