Rigorous constraint on metallicity from our forward-modeling Bayesian inference method. Left: marginalized 1D/2D constraints on metallicity (﹩12+\mathrm{log}({\rm{O}}/{\rm{H}})﹩), nebular dust extinction (﹩{A}_{{\rm{V}}}^{{\rm{N}}}﹩), and dereddened Hβ flux (f_Hβ) based on the integrated emission line fluxes of object MACS 0416–ID 00955, presented in Table A1 (also see Figures 2 and 3). The parameter inference values shown at the top of each column are medians with 1σ uncertainties drawn from the [16th, 50th, 84th] percentiles, marked by the vertical dashed lines in the 1D histograms. Green corresponds to the inference from the actual observed emission line fluxes where Hβ is detected at S/N ∼ 11 (i.e., ﹩{f}_{{\rm{H}}\beta }^{\mathrm{obs}}=6.88\pm 0.60﹩). Blue results are derived with the uncertainty of Hβ artificially increased by a factor of 10 (i.e., S/N ∼ 1.1), and other emission line flux measurements are unchanged. The comparison between the green and blue results shows that although the constraints on ﹩{A}_{{\rm{V}}}^{{\rm{N}}}﹩ and f_Hβ are severely worsened by the decrease in S/N of Hβ, the inference on ﹩12+\mathrm{log}({\rm{O}}/{\rm{H}})﹩ remains largely unchanged. This comparison thereby testifies that our forward-modeling Bayesian inference of metallicity does not require high-S/N detection of Hβ. Right: histograms of median metallicities measured using our forward-modeling method in all individual Voronoi cells from our entire galaxy sample. The distribution of metallicity measurements is divided into three groups corresponding to three different ranges of Hβ-observed S/N in the corresponding Voronoi cells. The horizontal dotted line shows the flat prior of ﹩12+\mathrm{log}({\rm{O}}/{\rm{H}})﹩ ∈ [7.0, 9.3] used in our Bayesian inference. We demonstrate that the returned metallicity estimates do not simply revert to the prior, and there is no systematic offset in our metallicity inference, even in the low-S/N regime of Hβ.