Bonnie’s Abstract: Presently a > 3σ tension exists between values of the Hubble constant H0 derived from analysis of fluctuations in the Cosmic Microwave Background by Planck and local measures of the expansion using calibrators of type Ia supernovae (SNe Ia). Calibration of SNe Ia requires a series of often subtle choices, and the need to preclude human bias is more compelling that ever before. We perform a blind reanalysis of Riess et al. (2011) to measure the H0 using SNe Ia from CfA3 and LOSS, calibrated by Cepheid variables and the geometric distance to NGC 4258. We obscure the value of H0 throughout our analysis and the referee process. Our analysis departs from that of Riess et al. (2011) by incorporating the covariance matrix method adopted in SNLS and JLA to quantify SN Ia systematics, and includes a simultaneous fit of all SN Ia and Cepheid data using MultiNest, thus capturing the full dependences and interactions of all parameters. The measured value of H0 is yet to be revealed; however, we find a relative uncertainty in H0 marginally larger than the errors in Riess et al. (2011) and in the Efstathiou (2014) reanalysis. Our error budget is dominated by statistical noise in the supernova data, whilst the systematic contribution to the error in H0 is dominated by uncertainties in the host galaxy mass dependence and Malmquist bias.
Samuel’s Abstract: With systematic and calibration uncertainty now on equal footing with statistical uncertainty for supernova cosmology, supernova analysis methodologies need improved techniques, rather than larger data sets, to achieve most precise and robust cosmological constraints. Bayesian methods offer one solution to this problem. After giving an overview of current and in-development analysis methodologies for supernova cosmology, I will detail my current work on Hierarchical Bayesian modelling. Specifically, the big question in supernova cosmology is how to accurate model biases and selection effects, and their non-analytic nature has driven a large push towards forward modelling. I present a new method to calculate biases that has been integrated within a Bayesian framework, without resorting to lower-dimension functional approximations.
