ModDS Documentation¶
ModDS is a code for specifying and sampling from Bayesian models of dark matter halos. It uses colossus for physical halo models and emcee for Markov chain Monte Carlo sampling.
The Parameter class keeps track of the prior distributions and parameter transformations for free parameters in the model. Prior probability distributions can either be taken from those in scipy.stats or constructed by hand.
The MeasurementModel class wraps up a parameterized halo mass profile (a subclass of colossus.halo.profile_base.HaloDensityProfile), a list of free Parameter instances, and a set of observables (e.g., density, mass or surface density as a function of radius). Once created, a MeasurementModel instance is callable, and maps a list of parameter values to a log posterior probability.
The h5io module contains utilities for managing hdf5 files. In particular, it provides the HDF5Backend class, which manages a saved model and the results of sampling from the model.
The installed modds script wraps all this together in a nice command line interface.