The next generation of radio telescopes may include space-based as well as ground-based experiments. In particular, the far side of the Moon, above the Earth's ionosphere and shielded from RFI, yields the best location for observations of the global 21-cm signal from the cosmic dark ages to the epoch of reionization. A SMEX mission, the Dark Ages Radio Explorer (DARE) has been proposed to NASA to exploit this unique environment. DARE would provide our first probe of the epoch of the very first stars, as well as providing complementary information to power spectrum measurements (and observations in other wavebands) at lower redshifts.

The exact shape of the sky-averaged 21-cm signal is unknown, however, which causes difficulties when trying to extract it from observations which also contain bright, astrophysical foreground emission. The precise form of the foregrounds is also unknown, creating a delicate signal separation problem which we approach via a Bayesian model selection methodology. I will discuss what information about the 21-cm signal, and the properties of the first objects, is retained in this procedure. I will go on to show how to maximise it by choosing appropriate signal models, and representing them with physically meaningful parametrizations.