discrete models of AGN

Elisabeta Lusso (MPIA), Joe Hennawi (MPIA), and X. Prochaska (UCSC) discussed physical models of AGN with me at lunch today. They are modeling each AGN as a linear combination of warm/cold dust emission, a dusty torus, a stellar population, and an accretion disk, attenuated by two extinctions, one for the stars and one for the AGN. For each of the four physical components they have a discrete set of possible models (a set of "archetypes" in my lingo). So the model parameters are four amplitudes, two extinction amplitudes, and four integers. But if they want to marginalize out most of this—and they do—they need to put priors over those integers. That is, they need four huge lists of probabilities, each of which sums to unity. I noted to them that we have done this very thing in our star–galaxy paper and told them I would write down some math in the next week or two. They want to run on every known AGN, using every extant data set, so that's my kind of problem!