Basics of Bayesian inference and parameter estimation

To use redback we assume some level of familiarity with Bayesian inference and model fitting. However, if this is not the case, bilby provides a basic demonstration of Bayesian inference and how it is implemented in bilby. An example for a basic problem of fitting a line is available in the bilby documentation.

redback workflow

In redback, we make this process homogenous specifically for fitting electromagnetic transients. The redback workflow for fitting is:

  • Download the data from a public catalog, or provide your own data. Or simulate it.

  • Load the data into a homogenous transient object, which does the necessary processing and provides simple way to plot data.

  • The user then specifies a model (either already implemented in redback or their own function).

  • Write a prior or use the default priors.

  • Specify a sampler and sampler settings as in bilby

  • Fit model!

  • The fit returns a homogenous result object, which can be used for further diagnostics, and provides a simple way to plot the fit.

More advanced fitting functionality

  • The likelihood is by default set by the type of transient/data used, the more advanced users can provide their own or use more complicated likelihoods implemented in redback.

  • Modify the physics of a transient model by passing in different class constructors (dependency injections).

  • Place constraints on priors if necessary.

  • Joint analysis.

  • Reweighting.

  • Hierarchical inference.