Question

I have language data containing over 200 languages with some missing values coded as ” (0 length characters).

I would like to compress this using factor to code main languages, and all others as ‘other language’ while the ” is coded as ‘(missing)’ showing up at the end of the string.

My plan is this:

lanfmt <- list(
  lev = c(prime <- c('English', 'Russian', 'Urdu'), diff <- setdiff(levels(lan), c(prime, '')), ''),
  lab = c(prime, diff, '')
)

table(factor(lan, lanfmt$levels, lanfmt$labels)

but R doesn’t like many-to-one formats of factors. How do I aggregate into a single category?

EDIT:

I decided a good solution, using lanfmt as described above is the following:

table(lanfmt$lab[match(lang, lanfmt$lev)])

It’s not as elegant, but it works in a pinch.

Answer 1

I think you should convert your factors to character, edit them and then order them. Maybe something like this would help (lan being the language vector of your list / data frame):

lan <- c("English", "Russian", "Urdu", "", "Indonesian")
lan <- factor(lan)
prime <- c("English", "Russian", "Urdu", "missing")
missing <- ""

lan <- as.character(lan)
lan[lan %in% missing] <- "missing"

lan[!lan %in% prime] <- "other language"
lan <- factor(lan)
lan
[1] English        Russian        Urdu           missing       
[5] other language
Levels: English missing other language Russian Urdu

After that you can order your languages

order <- c("English", "Russian", "Urdu", "other language", "missing")
lan <- ordered(lan, order)
dt <- data.frame(lan, stuff=rnorm(5,4,1))
dt[with(dt, order(lan)),]

             lan    stuff
1        English 4.212460
2        Russian 3.681616
3           Urdu 3.409838
5 other language 3.304108
4        missing 3.938468