Do we need a way to easily convert pre-calculated statistics into an ARD format?

[ddsjoberg]

Sometimes it will jsut be easier to calculate the quantity you need rather than wrapping it up in a function and processing it through existing ard_*() functions. In the example below, we're taking a pre-calculated dispersion statistic from a negative binomial regression model and adding it to an ARD.

Would this be helpful?

Something like this?

devtools::load_all("/Users/sjobergd/Documents/GitHub/cards")
#> ℹ Loading cards

ard_literal <- function(stats, variable, context = "literal") {
  # check inputs ---------------------------------------------------------------
  set_cli_abort_call()
  check_named_list_and_vector_elements(stats)
  check_string(variable)
  check_string(context)
  
  # build data frame for calculaton --------------------------------------------
  dplyr::tibble("{variable}" := TRUE) |> 
    ard_continuous(
      variables = all_of(variable),
      statistic = everything() ~ list(identity = \(x) stats)
    ) |> 
    dplyr::mutate(
      context = .env$context
    )
}

ard_literal(list(dispersion = 1.1), variable = "age", context = "negative binomial regression")
#> {cards} data frame: 1 x 8
#>   variable   context  stat_name stat_label stat fmt_fn
#> 1      age negative… dispersion  dispersi…  1.1      1
#> ℹ 2 more variables: warning, error

^{Created on 2025-01-22 with reprex v2.1.1}

[Melkiades]

I think this can be useful for tests. I do not know if users will be adopting it though, also because it applies only to one variable (no groups right?)

[ddsjoberg]

I think this can be useful for tests. I do not know if users will be adopting it though, also because it applies only to one variable (no groups right?)

We could add groups to the inputs, but if the results are pre-calculated I imagined grouping wouldn't be necessary...I can't think of a use case