Create specification object for ADaM data sets of type 'bds'

Given a file containing a bds data set (e.g. adlb or advs), adam_spec_bds() will create a specification object for use in build_bds() to prepare the data to be used in machine learning. The main task is to collect the key columns for reshaping the data into wide format and prepare the data filter.

Usage

adam_spec_bds(
  file = NULL,
  data = NULL,
  id = "USUBJID",
  param = NULL,
  label = NULL,
  unit = NULL,
  time = NULL,
  value = NULL,
  filter = NULL,
  attach_data = FALSE,
  domain = NULL
)

Arguments

file: the path of the sas(7bdat) or rds file to process, ignored if data is provided
data: tibble with the data in bds format for which the specification is created
id: name of id column to be kept and used for merge of data sets
param: name of the column that identifies the parameter. Defaults to NULL, will be guessed if not set (see Details).
label: name of the column that gives column labels. Defaults to NULL.
unit: Defaults to NULL, will be guessed if not set (see Details).
time: Defaults to NULL, will be guessed if not set (see Details).
value: Defaults to NULL, will be guessed if not set (see Details).
filter: character vector of filters to be applied to the bds data set. Individual filters will only be considered if the resulting data set has positive number of rows. Defaults to NULL.
attach_data: boolean. Attach the imported raw data.
domain: character string to be included in dictionary. Automatically derived for standard ADaM data sets. If not set for data provided, dictionary entry will be 'custom'.

Value

A list containing the following

file, md5: the name and md5 checksum, resp., of the file the generated spec is based upon
data: the raw data set if attach_data, NULL otherwise
data_info: a list containing the number of subjects nsubj and columns ncol in the data after applying filter
type: character string bds, generally giving the type of ADaM data set processed (adsl/bds/occds)
filter: subset of filter that yields valid and non-empty result when applied individually (using check_filter())
id: passing unchanged input
param, label, value, unit, time: names of the key columns to be used in build_bds() for reshaping
spec_id: character string, generally the name of the domain
dict: a tibble with unique combinations within the param and label column (if present in the data set) to be used as a data dictionary
dupl_ctrl: a list of length 2 with parameters values_fn and arrange that are passed to build_bds() to handle pivoting for duplicated values. Both default to NULL.

Details

Values for arguments param, label, unit, time and value will be guessed if not provided. Guess will be the first of the following options that matches a column name (exact match).

param: PARAMCD
label: PARAM
time: AVISIT, AVISITN, VISIT, VISITN
value: AVAL, AVALC
unit: AVALU

Function will escape if one of param or value are neither provided nor can be guessed. The other columns are optional.

Authors

Maike Ahrens (ahrensmaike), Sebastian Voss (svoss09)