forcats

Package forcats contain tools for working with EDAV - Categorical Data (R Type#Factors)

Recode Factor Levels

To recode the levels, do not use levels(x) <- c("A", "B"); use fct_recode() in forcats instead.

library(forcats)
x <- factor(c("G234", "G452", "G136"))  
y <- fct_recode(x, Physics = "G234", Math = "G452", Chemistry = "G136")  
y

Reorder Factor Levels

• fct_inorder(): by the order in which they first appear

• fct_infreq(): by number of observations with each level (largest first)

  • useful for unbinned data

• fct_inseq(): by numeric value of level.

• fct_relevel(x, level1, level2, after = 4): by hand

  • without after, you are putting levels to the beginning
  • with after = Inf, you are putting levels to the last

• fct_rev(): reverse the current order

• fct_reorder(x,y): by sorting along another variable

  • works with .desc = False

• fct_reorder2(x,y,z): by sorting along another 2 variables

• All the above functions return new factors.

Include NA

When a factor contains NAs, NAs will not form a level. And when plotting, NA always orders first (in the top-down graph, and last in the left-right graphs). We can use fct_explicit_na(x, "NA") to make NA a real level with the name "NA". Then we can reorder the levels including "NA"

df <- data.frame(temperature = factor(c("cold", "warm", "hot", NA)),
                 count = c(15, 5, 22, 12))

df %>%
  mutate(temperature = fct_explicit_na(temperature, "NA") %>% # try comment this and the following lines
      fct_relevel("NA", "hot", "warm", "cold")) %>%
  ggplot(aes(x = temperature, y = count)) +
  geom_col() +
  coord_flip()

Lumping

• fct_lump_*: lump together factor levels into "other"
  • fct_lump_min(): lumps levels that appear fewer than min
    times
    • cc |> mutate(continent_new = fct_lump_min(continent, 30)) puts all continents with less than 30 countries (items) into the "Other" category (level)
  • fct_lump_prop(): lumps levels that appear in fewer prop * n times, where n is the size of the data
  • fct_lump_n(): lumps all levels except for the n most frequent (or least frequent if n < 0)
  • fct_lump_lowfreq(): lumps together the least frequent
    levels, ensuring that "other" is still the smallest level

by zcysxy