Revisiting the Local Star-Forming Galaxies Observed in the HETDEX Pilot Survey

TL;DR

This study reanalyzed local star-forming galaxies from the HETDEX Pilot Survey, improving metallicity estimates, revealing overestimated star-formation rates in previous work, and emphasizing the importance of convergence monitoring in MCMC analyses.

Contribution

It introduces a careful Bayesian reanalysis of HETDEX data, highlighting the impact of convergence and calibration uncertainties on galaxy property estimates.

Findings

New metallicity estimates are higher by ~2-sigma than previous values.

Star-formation rates were overestimated by up to a factor of five.

Proper convergence monitoring is crucial for reliable MCMC results.

Abstract

I have reanalyzed the data obtained for local ($z<0.15$) star-forming galaxies during the pilot survey for the Hobby-Eberly Telescope Dark Energy Experiment (HETDEX)---called the HETDEX Pilot Survey (HPS)---which uses an integral-field-unit spectrograph and covers $\sim3500-5800$ $\unicode{x212B}$ at $\sim5$ $\unicode{x212B}$ resolution. I have newly determined the gas metallicities, $12 + \text{log(O/H)}$, following the Bayesian analysis scheme of the previous study, but dealing carefully with the uncertainty of strong-line calibration, performing reproducibility tests with mock data, and monitoring the convergence of the Markov-Chain Monte Carlo (MCMC) sampling. From the mock-data tests, I found that the nebular emission-line color excess $E(B-V)$ can be overestimated by as much as 2-$\sigma$ or more, although the metallicity can recover the input value to within 1-$\sigma$. The new metallicity estimates on the HPS data are from well-converged MCMC samples (effective sample sizes $>$ 2000), and they are higher than the previous estimates by $\sim$2-$\sigma$. Using the HPS data, I also showed that the MCMC sampling can have the statistical accuracy as poor as the one near the iteration start if done without convergence monitoring. The overestimation of $E(B-V)$ indicates the overestimation of the star-formation rates (SFRs) in the previous study, which can be as much as a factor of five. This finding undermines the previous suggestion of a hitherto-unknown galaxy population based on the locations of galaxies in the mass-SFR plane. I found that the independent determination of $E(B-V)$ using either H${\beta}$-H${\gamma}$ or H${\alpha}$-H${\beta}$ line pair is ideal for the analysis of forthcoming HETDEX data, but it requires additional cost.