Why are my factor levels turning to numbers with dplyr::mutate()?

Question

I'm trying to recode some factor levels using dplyr::mutate(), so that all "MOT" and "CHI" factor levels remain as they are, and all others turn to "OTH".

My data looks like this:

subj | speaker | word

1        MOT     apple
1        CHI     baby
1        SI1     baby
2        CHI     dog
2        CHI     cat
2        FAT     cat

And I want it to looks like this:

subj | speaker | word

1        MOT     apple
1        CHI     baby
1        OTH     baby
2        CHI     dog
2        CHI     cat
2        OTH     cat

My code looks like this:

new.df <- data %>%
  dplyr::select(subj, speaker) %>%
        mutate(speaker = factor(speaker),
         speaker = ifelse(speaker %in% c("CHI", "MOT"), speaker, "OTH")) 

However, when I run this script some of the levels are converted to "OTH" but all of the others change to numbers. This happens regardless of whether I include the factor(speaker) command:

subj | speaker | word

1        175     apple
1        86      baby
1        OTH     baby
2        86      dog
2        86      cat
2        OTH     cat

It is changing the factors to "OTH" ok but the TRUE element of my ifelse code appears to be the problem.

Many thanks!

Answer 1

Why I am not sure exactly but if you read the help menu for factor: ?factor it mentions first converting to as.character.

new.df <- data %>%
dplyr::select(subj, speaker) %>%
    mutate(speaker = factor(speaker),
     speaker = ifelse(speaker %in% c("CHI", "MOT"), as.character(speaker), "OTH"))