Skip to contents

The ADaM datasets as a summary for an AI chat

Source: R/ai_adam.R
tfl_ai_adam_summary.Rd

What an AI chat needs to know of the study's data to draft its specs, and nothing more: per dataset its label and size, and per variable its name, label and type, with the values of a character or factor variable that has few of them (SEX, AGEGR1, the arms, PARAMCD ...). No record is written: identifiers, dates and times never have their values listed. Paste the text into the chat; never attach the datasets themselves.

Usage

tfl_ai_adam_summary(path, datasets = NULL, max_levels = 15L, file = NULL)

Arguments

path

A folder of ADaM files (.xpt, also .sas7bdat, .rds), one file, or a named list of data frames. .xpt and .sas7bdat need the haven package.

datasets

The datasets to keep ("ADSL", "ADAE" ...); NULL for all.

max_levels

A variable's values are listed when it has this many at most (PARAMCD, PARCAT, AVISIT, the grades: four times as many).

file

A file to write the text to (UTF-8), or NULL.

Value

A tfl_ai_adam: a list with text (one string, Markdown), variables (one row a variable: dataset, dataset_label, variable, label, kind, levels – | between the values – as tfl_adam_metadata() gives it) and datasets (dataset, label, records, variables). format() gives the text.

Details

The summary's variables frame is what tfl_ai_context() takes as datasets: tfl_ai_check() then checks every dataset and variable an answer names against it.

See also

tfl_ai_read_shell(), tfl_ai_context().

Examples

s <- tfl_ai_adam_summary(list(adsl = tflspec::adsl, adae = tflspec::adae))
cat(substr(format(s), 1, 700))
#> # ADaM datasets: variables only
#> 
#> Each line: `NAME (type) Label: values`. The values are listed for a character variable with few of them; no record is shown.
#> 
#> ## ADSL (200 records, 45 variables)
#> STUDYID (char) Study Identifier
#> USUBJID (char) Unique Subject Identifier
#> SUBJID (char) Subject Identifier for the Study
#> SITEID (char) Study Site Identifier
#> COUNTRY (char) Country: CAN | DEU | ESP | FRA | GBR | JPN | KOR | USA
#> REGION1 (char) Geographic Region 1: ASIA | EUROPE | NORTH AMERICA
#> AGE (num) Age
#> AGEU (char) Age Units: YEARS
#> AGEGR1 (char) Pooled Age Group 1: <65 | >=65
#> AGEGR1N (num) Pooled Age Group 1 (N)
#> SEX (char) Sex: F | M
#> RACE (char) Race: ASIAN | BLACK OR AFRICAN AMERICAN | OTHER | WHIT
head(s$variables)
#>   dataset dataset_label variable                            label      kind
#> 1    ADSL          <NA>  STUDYID                 Study Identifier character
#> 2    ADSL          <NA>  USUBJID        Unique Subject Identifier character
#> 3    ADSL          <NA>   SUBJID Subject Identifier for the Study character
#> 4    ADSL          <NA>   SITEID            Study Site Identifier character
#> 5    ADSL          <NA>  COUNTRY                          Country character
#> 6    ADSL          <NA>  REGION1              Geographic Region 1 character
#>                                          levels
#> 1                                          <NA>
#> 2                                          <NA>
#> 3                                          <NA>
#> 4                                          <NA>
#> 5 CAN | DEU | ESP | FRA | GBR | JPN | KOR | USA
#> 6                 ASIA | EUROPE | NORTH AMERICA
# \donttest{
if (requireNamespace("haven", quietly = TRUE)) {
  d <- file.path(tempdir(), "adam")
  dir.create(d, showWarnings = FALSE)
  haven::write_xpt(tflspec::adsl, file.path(d, "adsl.xpt"))
  tfl_ai_adam_summary(d)
}
#> <tfl_ai_adam: 1 dataset(s), 45 variables, 2608 characters>
#>   ADSL (200 records, 45 variables)
#> format(x) gives the text to paste.
# }