skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: The overlooked potential of Generalized Linear Models in astronomy-II: Gamma regression and photometric redshifts

; ORCiD logo; ; ; ;
Publication Date:
Sponsoring Org.:
OSTI Identifier:
Resource Type:
Journal Article: Publisher's Accepted Manuscript
Journal Name:
Astronomy and Computing
Additional Journal Information:
Journal Volume: 10; Journal Issue: C; Related Information: CHORUS Timestamp: 2017-05-12 09:05:12; Journal ID: ISSN 2213-1337
Country of Publication:

Citation Formats

Elliott, J., de Souza, R. S., Krone-Martins, A., Cameron, E., Ishida, E. E. O., and Hilbe, J.. The overlooked potential of Generalized Linear Models in astronomy-II: Gamma regression and photometric redshifts. Netherlands: N. p., 2015. Web. doi:10.1016/j.ascom.2015.01.002.
Elliott, J., de Souza, R. S., Krone-Martins, A., Cameron, E., Ishida, E. E. O., & Hilbe, J.. The overlooked potential of Generalized Linear Models in astronomy-II: Gamma regression and photometric redshifts. Netherlands. doi:10.1016/j.ascom.2015.01.002.
Elliott, J., de Souza, R. S., Krone-Martins, A., Cameron, E., Ishida, E. E. O., and Hilbe, J.. Wed . "The overlooked potential of Generalized Linear Models in astronomy-II: Gamma regression and photometric redshifts". Netherlands. doi:10.1016/j.ascom.2015.01.002.
title = {The overlooked potential of Generalized Linear Models in astronomy-II: Gamma regression and photometric redshifts},
author = {Elliott, J. and de Souza, R. S. and Krone-Martins, A. and Cameron, E. and Ishida, E. E. O. and Hilbe, J.},
abstractNote = {},
doi = {10.1016/j.ascom.2015.01.002},
journal = {Astronomy and Computing},
number = C,
volume = 10,
place = {Netherlands},
year = {Wed Apr 01 00:00:00 EDT 2015},
month = {Wed Apr 01 00:00:00 EDT 2015}

Journal Article:
Free Publicly Available Full Text
Publisher's Version of Record at 10.1016/j.ascom.2015.01.002

Citation Metrics:
Cited by: 9works
Citation information provided by
Web of Science

Save / Share:
  • We present stellar locus regression (SLR), a method of directly adjusting the instrumental broadband optical colors of stars to bring them into accord with a universal stellar color-color locus, producing accurately calibrated colors for both stars and galaxies. This is achieved without first establishing individual zero points for each passband, and can be performed in real-time at the telescope. We demonstrate how SLR naturally makes one wholesale correction for differences in instrumental response, for atmospheric transparency, for atmospheric extinction, and for Galactic extinction. We perform an example SLR treatment of Sloan Digital Sky Survey data over a wide range ofmore » Galactic dust values and independently recover the direction and magnitude of the canonical Galactic reddening vector with 14-18 mmag rms uncertainties. We then isolate the effect of atmospheric extinction, showing that SLR accounts for this and returns precise colors over a wide range of air mass, with 5-14 mmag rms residuals. We demonstrate that SLR-corrected colors are sufficiently accurate to allow photometric redshift estimates for galaxy clusters (using red sequence galaxies) with an uncertainty {sigma}(z)/(1 + z) = 0.6% per cluster for redshifts 0.09 < z < 0.25. Finally, we identify our objects in the 2MASS all-sky catalog, and produce i-band zero points typically accurate to 18 mmag using only SLR. We offer open-source access to our IDL routines, validated and verified for the implementation of this technique, at« less
  • Based on the ultraviolet to far-infrared photometry already compiled and presented in a companion paper (Paper I), we present a detailed spectral energy distribution (SED) analysis of nearly 80,000 IRAC 3.6 + 4.5 {mu}m selected galaxies in the Extended Groth Strip. We estimate photometric redshifts, stellar masses, and star formation rates (SFRs) separately for each galaxy in this large sample. The catalog includes 76,936 sources with [3.6] {<=} 23.75 (85% completeness level of the IRAC survey) over 0.48 deg{sup 2}. The typical photometric redshift accuracy is {Delta}z/(1 + z) = 0.034, with a catastrophic outlier fraction of just 2%. Wemore » quantify the systematics introduced by the use of different stellar population synthesis libraries and initial mass functions in the calculation of stellar masses. We find systematic offsets ranging from 0.1 to 0.4 dex, with a typical scatter of 0.3 dex. We also provide UV- and IR-based SFRs for all sample galaxies, based on several sets of dust emission templates and SFR indicators. We evaluate the systematic differences and goodness of the different SFR estimations using the deep FIDEL 70 {mu}m data available in the Extended Groth Strip. Typical random uncertainties of the IR-bases SFRs are a factor of two, with non-negligible systematic effects at z {approx}> 1.5 observed when only MIPS 24 {mu}m data are available. All data products (SEDs, postage stamps from imaging data, and different estimations of the photometric redshifts, stellar masses, and SFRs of each galaxy) described in this and the companion paper are publicly available, and they can be accessed through our the Web interface utility Rainbow-navigator.« less
  • This paper proposes an alternative analysis of the statistically significant blood pressure/blood lead relationship reported for males, ages 12 to 74, based on data from the second National Health and Nutrition Examination Survey. Because of the substantial decline, both for blood lead levels and blood pressures, during the 4-year survey period, there is considerable interest in the extent to which this association can be attributed to concurrent secular trends. The statistical methods illustrate the use of a randomization model-based approach to testing the statistical significance of the partial correlation between blood lead level and diastolic blood pressure, adjusting for age,more » body mass index, and the 64 sampling sites. The resulting analyses confirm that the significant linear association between blood lead levels and diastolic blood pressures cannot be dismissed as a spurious association due to concurrent secular trends in the two variables across the 4-year survey period. In a conservative approach to this investigation, a randomization model-based test statistic, using the actual level of the natural log of blood lead and diastolic blood pressure, remained statistically significant at the 5% level, even when averaging the association across 478 subgroups formed by the cross-classification of age, body mass index, and the 64 sampling sites.« less
  • Following our previous work, which related generic features in the sky-averaged (global) 21-cm signal to properties of the intergalactic medium, we now investigate the prospects for constraining a simple galaxy formation model with current and near-future experiments. Markov-Chain Monte Carlo fits to our synthetic data set, which includes a realistic galactic foreground, a plausible model for the signal, and noise consistent with 100 hr of integration by an ideal instrument, suggest that a simple four-parameter model that links the production rate of Lyα, Lyman-continuum, and X-ray photons to the growth rate of dark matter halos can be well-constrained (to ∼0.1more » dex in each dimension) so long as all three spectral features expected to occur between 40 ≲ ν/MHz ≲ 120 are detected. Several important conclusions follow naturally from this basic numerical result, namely that measurements of the global 21-cm signal can in principle (i) identify the characteristic halo mass threshold for star formation at all redshifts z ≳ 15, (ii) extend z ≲ 4 upper limits on the normalization of the X-ray luminosity star formation rate (L{sub X}–SFR) relation out to z ∼ 20, and (iii) provide joint constraints on stellar spectra and the escape fraction of ionizing radiation at z ∼ 12. Though our approach is general, the importance of a broadband measurement renders our findings most relevant to the proposed Dark Ages Radio Explorer, which will have a clean view of the global 21-cm signal from ∼40 to 120 MHz from its vantage point above the radio-quiet, ionosphere-free lunar far-side.« less
  • The 3D-HST and CANDELS programs have provided WFC3 and ACS spectroscopy and photometry over ≈900 arcmin{sup 2} in five fields: AEGIS, COSMOS, GOODS-North, GOODS-South, and the UKIDSS UDS field. All these fields have a wealth of publicly available imaging data sets in addition to the Hubble Space Telescope (HST) data, which makes it possible to construct the spectral energy distributions (SEDs) of objects over a wide wavelength range. In this paper we describe a photometric analysis of the CANDELS and 3D-HST HST imaging and the ancillary imaging data at wavelengths 0.3-8 μm. Objects were selected in the WFC3 near-IR bands,more » and their SEDs were determined by carefully taking the effects of the point-spread function in each observation into account. A total of 147 distinct imaging data sets were used in the analysis. The photometry is made available in the form of six catalogs: one for each field, as well as a master catalog containing all objects in the entire survey. We also provide derived data products: photometric redshifts, determined with the EAZY code, and stellar population parameters determined with the FAST code. We make all the imaging data that were used in the analysis available, including our reductions of the WFC3 imaging in all five fields. 3D-HST is a spectroscopic survey with the WFC3 and ACS grisms, and the photometric catalogs presented here constitute a necessary first step in the analysis of these grism data. All the data presented in this paper are available through the 3D-HST Web site (« less