On the 30th of March, AIMS hosted the second seminar of the Joint Western Cape Astronomy/Cosmology Seminar series, which was a talk by Roberto Trotta on Bayesian cosmological model building. Roberto began by introducing the cosmic microwave background radiation (CMB) and describing how WMAP has measured the CMB with great accuracy. He also explained the angular power spectrum and showed how curvature can affect the power spectrum but that this effect is degenerate with dark energy and thus curvature cannot be constrained from the power spectrum alone.
This degeneracy can be broken by introducing additional data such as supernovae Ia or baryon acoustic oscillations (BAO) data. The problem of curvature is a good example for Bayesian model comparison, which makes use of the Bayesian evidence: a quantity that gives as indication of how likely a model is, given the data. The Jeffrey’s scale of ln B (where B is the Bayesian evidence) was introduced as a way of quantifying whether or not one model is significantly better than another.
Roberto then explained how you can perform model selection on the three possibilities of curvature in the universe (flat, open and closed), using two different priors: the astronomer’s prior and the curvature prior. They found that the results are quite dependent on the prior, which is something to be wary of when performing model selection. He hopes to apply this to models of inflation in the future.