Set "don't know", "refused" and other missing codes to NA

In the data get_qes() returns, the codes that stand for a missing answer ("don't know", "refused", SPSS user-missing codes such as 8 or 9, -99 in qes2022) are kept as values, as the files deposit them. qes_missing() turns them into NA, or into tagged NA values that remember the reason, using the missing type the codebook records for each code (see qes_codebook(), column missing_codes).

Usage

qes_missing(
  x,
  variables = NULL,
  action = c("na", "tagged"),
  types = NULL,
  quiet = FALSE
)

Arguments

x: A data frame returned by get_qes() (its study is read from its attributes; a data frame that has lost them is an error of class qesR_error_no_provenance).
variables: Optional character vector of column names (exact match) to recode; NULL recodes every column.
action: "na" (default) sets the codes to NA; "tagged" sets them, in numeric columns, to haven::tagged_na() values whose tag is the letter of their type (other columns get a plain NA).
types: Optional character vector of missing types to recode. NULL (default) means every type except spoiled, not_selected, not_voted and not_registered.
quiet: If TRUE, suppress the message that counts the variables with no typed codes.

Value

x, returned visibly, with the codes recoded and the attribute qes_missing_log: a data frame with one row per recoded code and variable (variable, value, missing_type, n_set, the number of values set to NA). Labels and other attributes are kept.

Details

Missing types, and their tag for action = "tagged": dk (d), refused (r), dk_refused (b, one code for both), no_answer (o), not_selected (s, an option not ticked in a multiple-choice question), inapplicable (i), not_voted (v), spoiled (p), ineligible (e), not_registered (g), not_in_wave (w), not_mappable (m) and user_na (u, a code the SPSS file declares missing without saying why). A code has one type. The codes of each study were typed from their labels, by hand; a variable whose codebook types none of its codes, and whose file declares none, is left unchanged, and a message counts those variables. Some files declare missing a code that is an answer ("Un autre parti", or "ne voterait pas/annulerait" in a voting-intention item): such a code has no missing type and is never changed. A declared code for "did not vote" or a spoiled ballot has the type not_voted or spoiled, so the default leaves it alone.

qes_missing() uses the codebook get_qes() attached to x, else the study's metadata shipped with qesR. It never downloads anything.

En français

qes_missing() remplace par NA les codes qui représentent une non-réponse (« ne sait pas », « refus », codes manquants déclarés dans le fichier SPSS, -99 dans qes2022), selon le type que le codebook attribue à chaque code. Avec action = "tagged", les colonnes numériques reçoivent des NA étiquetés (haven::tagged_na()) qui conservent la raison. Par défaut, les types spoiled, not_selected, not_voted et not_registered, qui sont des réponses, ne sont pas touchés. Un code déclaré manquant dans le fichier mais qui est une réponse (« Un autre parti », « ne voterait pas/annulerait » dans une intention de vote) n'a pas de type et n'est jamais modifié. La fonction ne télécharge rien.

Examples

demo <- get_qes("qes_demo", quiet = TRUE)
table(demo$Q19, useNA = "ifany")
#> 
#>  1  2  8  9 
#> 22 33  3  2 
clean <- qes_missing(demo, variables = c("Q19", "Q28"))
table(clean$Q19, useNA = "ifany")
#> 
#>    1    2 <NA> 
#>   22   33    5 
attr(clean, "qes_missing_log")
#>   variable value missing_type n_set
#> 1      Q19     8           dk     3
#> 2      Q19     9      refused     2
#> 3      Q28     8           dk     1
#> 4      Q28     9      refused     0

# keep the reason: "don't know" and "refused" become different NAs
tagged <- qes_missing(demo, variables = "Q19", action = "tagged")
table(haven::na_tag(tagged$Q19), useNA = "ifany")
#> 
#>    d    r <NA> 
#>    3    2   55