On the Impact of Cepheid Outliers on the Distance Ladder

Recent work by Efstathiou (2014) highlighted the importance of outliers in the period-luminosity (PL) relation of Cepheid data on the distance ladder. We present a statistical framework designed to address this difficulty, and apply it to the Cepheid data from the Milky Way (MW), the Large Magellanic Cloud (LMC), and the Riess et al. (2011) (hereafter R11) dataset. We consider two possible models of the outlier population in the R11 Cepheid dataset. One of these models exhibits tension between the PL relation of the R11 cepheids and the MW+LMC cepheids, while the other does not. We extend our models to adequately account for tension between the cepheid data sets when appropriate. Our outlier treatment has a significant impact on the distance scales to Supernovae hosts with Cepheid distances, increasing the uncertainty in these distances by a median factor of ~30%. We further find that our Cepheid outlier treatment translates into a modest, but non-negligible increase in the statistical uncertainty of H0, adding in quadrature 1.2 km/s/Mpc. Combined with the increased scatter in the Hubble diagram reported by Jones et al. (2015), we find H0=72.6+/-2.8 km/s/Mpc, corresponding to a 3.8% uncertainty in H0 from local measurements. This value is fully consistent with both the Planck and inverse-distance ladder H0 constraints.