Do people have any thoughts on how to output the result? Currently, it's just pickle dumped.

I've not used pickle much, is there a reason to use this over something hdf5?

I think we need/should have:

• A collection of posterior samples that can be read into a pandas data frame with the parameter names attached.
• The actual samples (nested samples and weights for nested samplers) in the same format.
• Any evidences we want along with associated errors.
• Injected parameters if any.

Other questions:

• Should parameters which are fixed be added as columns in the posterior. It's more storage-intensive but will make it easier to do some post-processing.

Thoughts?

@gregory.ashton @paul-lasky @moritz.huebner @rory-smith @eric.thrane