Future Version Ideas

[bstaton1]

postpack v0.5.1 is fully functional and documented. It will be good enough for now. Ideas for the future:

Warnings instead of Errors

• If an element of params is not found in post, it will trigger an error. Consider making this a warning instead. This would allow greater flexibility in analyzing output from multiple models using the same code.

Facilitate output summarization by user-defined groups

• Currently, params must be a character vector of regular expressions.
• If we allow params to be a named list where each list elements contains such a character vector of regular expressions, then the output could be returned in a named list, where the list elements represent main parameter groups (e.g., "states", "pop_dyn_params", "derived", "nuisance").
• Another function could take output in this format from post_summ() and return a data frame, with rows as nodes and columns as summary statistics. Additional columns could be used for group, param name, base name, indices. This could facilitate writing large model output for sharing critical summaries with non 'postpack' users or storing summaries from large simulations where-in it is unrealistic to save the 'mcmc.list' object created by each run.
• See the summarize_posterior() function in this code for an example of this in action

Create infrastructure for global 'postpack' options

• To control things like digits, auto_escape
• To specify a default 'mcmc.list' to use in all functions that accept a post argument. E.g., it might take the form of:

  postpack_options(default_post = post)
  post_summ("alpha")

  So if a user is only working with one 'mcmc.list' object, they don't have to keep typing its name each time they call a 'postpack' function.
• See the GlobalOptions package for what looks like a great way to handle this

Include NA values if element not touched by model

• If a JAGS model touches only node elements a[5], a[6], and a[7], nodes a[1:4] won't exist in the output.
• For structural reasons, it might be nice to have NA values for all summaries of nodes a[1:4]
• Including these would be an argument to post_summ()