Skip to contents

Every ard_*() of cards and cardx that is installed, with what a person reads to choose one: its category (the group it is listed under), a label (the heading) and a one-line description, and its shape – how it takes its input:

Usage

tfl_ard_functions(installed = TRUE)

Arguments

installed

TRUE (default) only the functions that can be called here (their package installed); FALSE every row of the catalog.

Value

A data frame: call, category, label, description, shape, replaced_by, offered, installed, in_catalog.

Details

  • variables: columns (by, variables, strata) of the data;

  • formula: a model formula (AVAL ~ TRT01P) and the data;

  • model: a fitted model, written in args (x = lm(..., data = data));

  • columns: named columns of its own (time, count; postbaseline);

  • wrapper: runs other analyses (ard_stack(), ard_strata(), ard_pairwise());

  • data: the data alone.

An analysis row names one of them as its method (cards::ard_summary), as it names any function; this list is the catalog that describes them. A function the catalog does not describe (one a newer cards adds) is listed too, with in_catalog = FALSE and its help page's title as its label. replaced_by names the new name of an old one. offered is FALSE for what a new analysis is not to choose: the old names and the functions not offered by design (below); a screen lists them only for an analysis that already names one.

The versions it is written for: cards >= 0.8.0 and cardx >= 0.3.1 (the oldest are tested, and every week what CRAN has now).

Not offered, by design (decided 2026-10-04)

Listed, but not for an analysis row:

  • the old names (ard_continuous(), ard_categorical(), ard_dichotomous(), ard_complex(), ard_categorical_max(), ard_emmeans_mean_difference()): the function in replaced_by gives the same result;

  • the survey design functions (ard_survey_svychisq(), ard_survey_svyranktest(), ard_survey_svyttest(), and the survey.design methods of the summaries): their input is a survey design object, not the data frame an analysis row reads;

  • ard_formals(): it records a function's argument values, a building block for writing an ARD function, not an analysis.

A company adds its own rows – its own ARD functions, other headings – with options(tflspec.ard_functions = <data frame>) (the same columns; a row with the same call replaces the built-in one).

See also

tfl_ard_args() for one function's arguments; tfl_ard_methods() for the keywords an analysis may name.

Examples

f <- tfl_ard_functions()
f[f$category == "Summaries", c("call", "label")]
#>                         call                          label
#> 1         cards::ard_summary             Summary statistics
#> 2        cards::ard_tabulate            Counts and percents
#> 3  cards::ard_tabulate_value             Count of one level
#> 4         cards::ard_missing                 Missing counts
#> 5   cards::ard_tabulate_rows                 Number of rows
#> 59     cards::ard_continuous  Summary statistics (old name)
#> 60    cards::ard_categorical Counts and percents (old name)
#> 61    cards::ard_dichotomous  Count of one level (old name)