2016/04/05 by Rafael Izbicki, Izbicki, Rafael, Ann B. Lee +3
Decision Sciences · Mathematics · #Advanced Statistical Methods and Models #Advanced Statistical Process Monitoring #Applications (stat.AP) #FOS: Computer and information sciences #FOS: Physical sciences #Instrumentation and Methods for Astrophysics (astro-ph.IM) #Statistical Methods and Inference
paper · pdf · doi:10.48550/arxiv.1604.01339
openalex publication_date 2016/04/05 · openalex created_date 2022/10/04 · openalex updated_date 2026/07/28
Redshift is a key quantity for inferring cosmological model parameters. In\nphotometric redshift estimation, cosmologists use the coarse data collected\nfrom the vast majority of galaxies to predict the redshift of individual\ngalaxies. To properly quantify the uncertainty in the predictions, however, one\nneeds to go beyond standard regression and instead estimate the full\nconditional density f(z|x) of a galaxy's redshift z given its photometric\ncovariates x. The problem is further complicated by selection bias: usually\nonly the rarest and brightest galaxies have known redshifts, and these galaxies\nhave characteristics and measured covariates that do not necessarily match\nthose of more numerous and dimmer galaxies of unknown redshift. Unfortunately,\nthere is not much research on how to best estimate complex multivariate\ndensities in such settings. Here we describe a general framework for properly\nconstructing and assessing nonparametric conditional density estimators under\nselection bias, and for combining two or more estimators for optimal\nperformance. We propose new improved photo-z estimators and illus- trate our\nmethods on data from the Sloan Data Sky Survey and an application to\ngalaxy-galaxy lensing. Although our main application is photo-z estimation, our\nmethods are relevant to any high-dimensional regression setting with\ncomplicated asymmetric and multimodal distributions in the response variable.\n