Skip to contents

Summarise item-level missings per person

summarise_missings.Rd

Counts the final missing classifications on item level and divides each count by the number of items assigned to that person in the completed design. Uses code_type; neither scores nor technical statuses are used to infer or change classifications.

Usage

summarise_missings(
  data,
  identifiers = c("group_id", "login_name", "login_code"),
  used_only = FALSE
)

Arguments

data

Data frame returned by complete_design() and enriched with add_item_id(). Requires unit_key, item_id, code_type, and all identifiers. Unmapped variables have a missing or blank item_id.

identifiers

Character vector of columns that jointly identify a person. All specified columns must exist and contain non-missing, non-blank values. Include an occasion identifier when summarising separate administrations of the same items.

used_only

Logical. Keep only people with id_used = TRUE? Defaults to FALSE, retaining all people in data. If present, id_used must be logical, non-missing, and constant within each person; it is returned in the result. It is required when used_only = TRUE.

Value

An ungrouped tibble (also a data frame) with one row per person, the identifiers, optional id_used, and n_items. Counts have prefix n_; proportions have prefix prop_ and range from 0 to 1:

  • missing: total of all MISSING_* types except MISSING_BY_DESIGN.

  • not_reached, by_omission, invalid_response, coding_impossible: the four standard missing types, always included even when absent.

  • Additional observed missing types use their full type as a suffix, converted to lowercase and made syntactically valid with make.names(), e.g. MISSING_TIMEOUT gives n_missing_timeout and prop_missing_timeout. Ambiguous generated names are an error.

  • no_coding, coding_incomplete, derive_pending, intended_incomplete: unresolved coding states, always included and excluded from the missing total.

  • unclassified: missing or blank code_type, also excluded from the missing total. Other non-missing types contribute to n_items only.

Details

An item is identified by unit_key and item_id together. Variables without an item mapping are excluded from item counts. More than one row for the same person and item is an error, even if the rows are identical. Resolve duplicate mappings or repeated administrations before summarising.

The denominator n_items includes all mapped items in the person's design, including missing responses and unresolved coding results. Combinations absent from the design are not added. Explicit MISSING_BY_DESIGN rows are excluded from both the denominator and missing counts. Retain all design rows before calling this function; response-only data underestimate the denominator and missings.

All people in the input are retained by default, including those without mapped items. People without eligible items receive zero counts and NA_real_ proportions. Empty input returns zero rows with the standard result columns. Person order follows first appearance in the input. Existing grouping is ignored.

See also

complete_design(), add_item_id()

Examples

data <- tibble::tibble(
  person = c("P1", "P1", "P1", "P2"),
  unit_key = "U1", item_id = c("01", "02", NA, "01"),
  code_type = c("FULL_CREDIT", "MISSING_NOT_REACHED",
                "MISSING_NOT_REACHED", "MISSING_BY_OMISSION"),
  id_used = TRUE
)
summarise_missings(data, identifiers = "person")
#> # A tibble: 2 × 23
#>   person id_used n_items n_missing prop_missing n_not_reached prop_not_reached
#>   <chr>  <lgl>     <int>     <int>        <dbl>         <int>            <dbl>
#> 1 P1     TRUE          2         1          0.5             1              0.5
#> 2 P2     TRUE          1         1          1               0              0  
#> # ℹ 16 more variables: n_by_omission <int>, prop_by_omission <dbl>,
#> #   n_invalid_response <int>, prop_invalid_response <dbl>,
#> #   n_coding_impossible <int>, prop_coding_impossible <dbl>, n_no_coding <int>,
#> #   prop_no_coding <dbl>, n_coding_incomplete <int>,
#> #   prop_coding_incomplete <dbl>, n_derive_pending <int>,
#> #   prop_derive_pending <dbl>, n_intended_incomplete <int>,
#> #   prop_intended_incomplete <dbl>, n_unclassified <int>, …