Convert Missing Values for the European Social Survey

ess.R

# Put both labels and values of the types of ESS missing values in a
# list to re-use later.
ESS_MISS <- list(
  not_applicable = list(pattern = "Not +.p+lic+able", value = 6),
  refused = list(pattern = "Refus..", value = 7),
  not_available = list(pattern = "(No .nswer|Not .vailable)",
                       value = 8),
  dont_know = list(pattern = "Don.?t .now", value = 9)
)

ess_missings <-
  function(x, missings = c("not_applicable", "refused",
                           "not_available", "dont_know")) {
  miss <- ESS_MISS[missings]
  miss_pattern <- str_c("(", map_chr(miss, "pattern"), ")", collapse = "|")
  if (is.factor(x)) {
    # If factor, drop old levels
    lvls <- levels(x)
    new_lvls <- lvls[!str_detect(names(lvls), miss_pattern)]
    factor(x, levels = new_lvls)
  } else if (is.character(x)) {
    # If character, set values to missing
    x[str_detect(x, miss_pattern)] <- NA
  } else if (is.labelled(x)) {
    # If haven::labelled, then set values to missing
    # And adjust labels
    lbls <- attr(x, "labels")
    na_lbls <- lbls[str_detect(names(lbls), miss_pattern)]
    new_lbls <- lbls[set_diff(names(lbls), names(na_lbls))]
    x[x %in% vals] <- NA
    x <- labelled(x, new_lbls)
  } else if (is.numeric(x)) {
    # If numeric value, then set values to missing
    # 1 = <10, 2 <100, 3 < 1000
    dig <- ceiling(log10(max(x, na.rm = TRUE)))
    # the rep() turns 6 to 6, 66, or 666 depending on the digits, etc.
    x[x %in% map_dbl(miss, ~ rep(.x[["value"]], dig))] <- NA
  } else {
    stop(x, " is an unsupported type: ", class(x), call. = FALSE)
  }
  x
}