Facilitate user passing in priors

[athowes]

Goal

As a part of a Bayesian analysis, users of this package should be able to easily pass in prior distributions they would like to use. We should also provide sensible default options.

Context

epidist interfaces via brms making it easy to help users pass in some priors. For other parameters (custom) then it's harder.

Required features

• Update epidist_priors to document the names of the parameters that users can pass in
• Update epidist::epidist to accept input of priors
• Possibly change the default brms priors if we think there is good reason to

Out of scope

• Likely out of scope to allow for passing in priors for custom parameters (not the brms default ones)

Related documents

• S3 structure refactor #70

[athowes]

I've been thinking about this and one thing we could do is like:

epidist_priors <- function(data, family, formula, ...) {
  default_prior <- brms::default_prior(
    object = formula,
    data = data,
    family = family
  )
  
  # Now replace parts of the default priors with user provided priors
  priors <- ...
  # Now warn user that "default brms prior used for parameter: ... pass in something to replace it..."
  return(priors)
}

What do you think @seabbs? Am playing with this at the moment.

[seabbs]

yeah could do it as the output and update the list using ModifyList

[athowes]

What do you mean by "as the output"?

The output of epidist_priors at the moment is being passed into epidist.

I'd imagine that the output is the list of priors, as required by brms.

I think that this should be the output, and then there should be a message about which default priors have been used.

Perhaps there could just be one message at the end to say how to change the priors.

Perhaps the messages could be set to be off by default somehow when passed in as default argument to epidist (don't want to spam user with too many messages, just want them to get this information if they care about changing the priors).

[seabbs]

epidist(..., priors = ModifyList(epidist_priors(), sigma = blah blah blah))`

could be a pattern