A Bayesian Estimate of the Primordial Helium Abundance

We introduce a new statistical method to estimate the primordial helium abundance, Y_p from observed abundances in a sample of galaxies which have experienced stellar helium enrichment. Rather than using linear regression on metal abundance we construct a likelihood function using a Bayesian prior, where the key assumption is that the true helium abundance must always exceed the primordial value. Using a sample of measurements compiled from the literature we find estimates of Y_p between 0.221 and 0.236, depending on the specific subsample and prior adopted, consistent with previous estimates either from a linear extrapolation of the helium abundance with respect to metallicity, or from the helium abundance of the lowest metallicity HII region, I Zw 18. We also find an upper limit which is insensitive to the specific subsample or prior, and estimate a model-independent bound Y_p < 0.243 at 95% confidence, favoring a low cosmic baryon density and a high primordial deuterium abundance. The main uncertainty is not the model of stellar enrichment but possible common systematic biases in the estimate of Y in each individual HII region.