Recode a worded rating scale to integers

Maps the worded answers of an ordinal rating question (e.g. "Very bad" ... "Very good") onto integers 1:length(levels). Matching is case-insensitive and tolerant of common synonyms supplied via synonyms, in place of a case_when(str_detect("Very bad") ...) recoding block.

Usage

recode_likert(
  x,
  levels = c("Very bad", "Bad", "Ok", "Good", "Very good"),
  synonyms = NULL
)

Arguments

x

A character (or factor) vector of worded answers.

levels

Character vector of the scale's answer wordings in ascending order. The first element maps to 1, the last to length(levels). Defaults to a 5-point bad-to-good scale.

synonyms

Optional named list mapping a canonical level (one of levels) to a character vector of alternative spellings that should map to the same integer. Matching is by case-insensitive substring.

Value

An integer vector the same length as x; unmatched values become NA.

Details

Matching happens in three passes, in order: (1) an exact, case-insensitive, trimmed match against levels; (2) a substring fallback, so prefixed exports like "4 - Good" still resolve; (3) any synonyms you supply. This makes the function robust to the small wording differences between survey tools (e.g. "Dissatisfied" vs "Bad"). Turning ratings into integers is the first step of ipm_model() and lets you average a scale with calc_summary().

The substring passes try the longest wording first, so a scale whose levels nest inside one another ("Likeable" inside "Very likeable", "Good" inside "Very good") resolves to the level the answer actually names. Without that, any answer the exact pass misses, such as one carrying an emoji or a stray character, would quietly land on the shorter level.

See also

nps_group(), ipm_model(), calc_summary().

Other recode: add_region(), bin_numeric(), clean_label(), drop_items(), ensure_numeric(), generation_scheme(), na_blank(), nps_group(), recode_age(), recode_generation(), recode_region(), split_multi()

Examples

recode_likert(c("Very bad", "Ok", "Good", "Very good"))
#> [1] 1 3 4 5

# substring fallback copes with numbered labels
recode_likert("4 - Good")
#> [1] 4

# nested wordings resolve to the level the answer names, not the shorter one
recode_likert(c("\U0001F642 Very good", "\U0001F610 Good"))
#> [1] 5 4

# map another tool's wording onto the same scale
recode_likert(c("Dissatisfied", "Satisfied"),
              synonyms = list(Bad = "Dissatisfied", Good = "Satisfied"),
              levels = c("Very bad", "Bad", "Ok", "Good", "Very good"))
#> [1] 2 4