Skip to main content
SearchLoginLogin or Signup

A Comparison of the Accuracy and Precision of DEEP2 Spectroscopic and Photometric Galaxy Redshift Measurements Using a Statistical Measurement Error Model

Presentation #235.06 in the session “Galaxies”.

Published onJan 11, 2021
A Comparison of the Accuracy and Precision of DEEP2 Spectroscopic and Photometric Galaxy Redshift Measurements Using a Statistical Measurement Error Model

All galaxy redshift z measurements are, not surprisingly, subject to measurement errors. Measurement errors have both systematic components referred to as bias or lack of accuracy and random components referred to as imprecision or lack of precision. Because imprecision is probabilistic, it is typically reported as a standard deviation. In general, different methods may exhibit differences in the size of the measurement unit - that is they have different scales which results in a scale bias. (Ignoring an existing scale bias will lead to an incorrect assessment of imprecision in addition to the bias.) Bias is always relative to a standard, if one is known (absolute accuracy), or if not, an arbitrarily chosen method (relative accuracy). Spectroscopy is considered the gold standard for measuring redshift but in many situations can be difficult to perform. Photometry often provides and easier way to determine redshift. Redshifts from the DEEP2 Galaxy Redshift Survey were used to compare bias and imprecision of both spectroscopic and 6 types of photometric redshift measurements. Two different sets of galaxies were used. One set of galaxies had non-replicated measurements using 6 photometric methods and a corresponding non-replicated spectroscopic measurement (see Figure 1). The other set had two independently replicated spectroscopic redshifts (see Figure 2). A measurement error model which required 3 latent variables (one to handle what was common to the photometric methods and the other two handle the spectroscopic measurements in each set) was used (see Figure 3). A structural model connected the photometric redshifts to the spectroscopic measurements. Full information maximum likelihood was estimated using the OpenMx R package and included bootstrapped estimates of confidence intervals for each of the 24 parameters in the model. The results showed that the spectroscopic imprecision was 0.009 while the imprecision of the photometric measurements was from 10.5 to as much as 18.1 times that of the spectroscopic redshifts. The scale bias of each photometric method ranged from 0.97 to 1.06 relative to the spectroscopic scale (which was equal to 1).

Comments
0
comment
No comments here