Inspect label repairs and divergences found against the PDF data dictionary

Statistics Canada's PUMF command files routinely ship truncated value and variable labels – hard cuts at 60 characters, dropped leading text, dropped interior text. The damage is upstream of the command files (SAS, SPSS and Stata carry byte-identical text), but the same survey's user guide contains a data dictionary with the full label text.

Usage

pumf_label_repairs(tbl, action = NULL)

Arguments

tbl

A lazy `dplyr::tbl()` returned by [get_pumf()].

action

Optional filter, e.g. `"repaired"` or `c("repaired", "filled")`.

Value

A tibble with columns `kind` (`"variable"`/`"code"`), `name`, `val`, `lang`, `label_command_file`, `label_pdf`, `action`, `reason` and `validation`. Zero rows when the survey ships no parseable PDF dictionary or nothing diverged.

Details

For surveys that ship such a guide, `canpumf` parses it during metadata preparation, validates it against the microdata using the per-code frequencies the guide prints, and then repairs labels the guide demonstrably extends. This function returns the ledger of what it found: every divergence between the command file and the guide, whether or not it was acted on.

`action` is one of:

`repaired`

The command-file label was replaced, because the guide's text extends it *and* the command-file label carries a damage signature: either it sits at the width the command file's labels were hard-cut to, or it is what a dropped prefix leaves behind (a strict suffix of the guide's text). A guide that merely words a label differently is `flagged` instead, as is one whose extra text is an annotation prepended to a label the command file has in full – recognised by the `Key: value` shape of the dropped text, which running prose lost to truncation does not have.

`filled`

The command file had no label at all; the guide supplied one.

`flagged`

Recorded but not acted on – the two simply differ, or the variable's frequencies contradicted the data file, or the guide documents a code the command file never declared. Read `reason` for which.

The `validation` column carries the variable's frequency-check status (`validated`, `continuous`, `unchecked`, `mismatch`, or `not documented`), so a repair corroborated against the microdata can be told apart from one the check simply could not reach. See [pumf_freq_validation()].

See also

[pumf_var_labels()], [pumf_freq_validation()]

Examples

# \donttest{
gss <- get_pumf("GSS", "Cycle 16 (2002)")
if (!is.null(gss)) {
  pumf_label_repairs(gss, action = "repaired")
  close_pumf(gss)
}
# }