How to detect all binary character columns (each column has different set of char. values) in a data frame and convert them to 1s and 0s all at once?

Question

Is there is a clean way to detect all binary character columns in a data frame and convert them to 1s and 0s all at once. For example a column that contains only "yes" and "no" values and a column that contains just "day" and "night" values, etc. would all be converted to 1s and 0s with the same piece of code that doesn't require me to specify the words "yes" "no" "day" "night" "good" "bad", the list goes on?

Answer 1

library(tidyverse)

# Create a dataset with two binary variables (x2 and x4)
sample_data <- tibble(x1 = rnorm(1000),
       x2 = rbinom(1000, 1, 0.5),
       x3 = rpois(1000, 5),
       x4 = sample(c("yes", "no"), 1000, replace = TRUE))

# Determine which variables have two levels and save them
binary_vars <- sample_data %>%  
  # This line calculates how many different values are present within each variable
  map_df(~ unique(.) %>% length()) %>% 
  # These lines just clean up the results
  gather() %>% 
  arrange(value) %>% 
  filter(value == 2) %>% 
  # This line pulls the variable names
  pull(key)

# Define a function to convert all binary variables to 1s and 0s
make_binary <- function(vct) {
  vct %>% 
    as_factor %>% 
    as.numeric() %>% 
    `-`(1)
}

# Mutate the relevant variables
sample_data %>%
  mutate_at(binary_vars,
            make_binary)

Answer 2

Here's a cut that lets you define your own set of binaries.

The premise is that the first value in the list is "1", the remaining entries are all 0. In this case I've done vectors of length 2, but it's feasible you could do more.

binaries <- list(
  c("yes", "no"), c("day", "night"), c("on", "off"), c("true", "false")
)

dat <- data.frame(
  v1 = c("yes", NA, NA),
  v2 = c("yes", "maybe", "no"),
  v3 = c("true", "false", NA),
  v4 = c("hello", "goodbye", NA),
  stringsAsFactors = FALSE
)

possibly_binary <- function(x, binaries, na.rm = TRUE) {
  if (na.rm) binaries <- lapply(binaries, c, NA)
  foundsomething <- sapply(binaries, function(b) all(x %in% b))
  if (any(foundsomething)) {
    one <- binaries[[ which(foundsomething)[1] ]][1]
    return(+(x == one))
  } else return(x)
}

Here it is in action. We control what NA does with the na.rm= argument to the function. If it is true, then NA is effectively added to each of the binaries vectors, though it will be kept as NA in the returned data.

dat
#     v1    v2    v3      v4
# 1  yes   yes  true   hello
# 2 <NA> maybe false goodbye
# 3 <NA>    no  <NA>    <NA>

dat[] <- lapply(dat, possibly_binary, binaries = binaries)
dat
#   v1    v2 v3      v4
# 1  1   yes  1   hello
# 2 NA maybe  0 goodbye
# 3 NA    no NA    <NA>