Custom post-processing

[athowes]

Goal

Add model post-processing functions to the epidist package which interface with the S3 refactor (#70)

Context

The epidist package now uses S3 classes by model (only one currently). It now has the capacity to fit models with multiple families as specified in epidist_family.

Required features

• An extract_samples function
• An extract_summaries function
• Plotting functions? Maybe out of scope for this issue

prep_obs <- epidist_prepare(obs, model = "latent_individual")
fit <- epidist(prep_obs)
samples <- extract_samples(fit, ...)
summaries <- extract_summaries(fit, ...)

Probably samples and summaries should be data.frame objects. Unclear exactly what they should contain.

Specifications

• Remains to fill

Out of scope

• Remains to fill

Uncertainties

• Quality of existing brms post-processing options and how much to direct users to rely on them
• Whether we should make the family attribute into a class
  • Could put classes for obvious things and let everything else go in default?
• Whether we should support particular families e.g. lognormal and gamma to a greater extent
• Whether we should restrict choice of family

[athowes]

I'm removing this issue as no longer relevant. I'll write a new post-processing meta issue.