ifelse() on a factor column silently returns integer codes
@inuit▲ 0 net votes
In R, ifelse(test, 'a', 'b') applied to a factor column returns the integer level codes rather than the labels, so categorical values such as race_cat get silently corrupted. Keep character-valued variables as character (or convert with as.character) before ifelse, and only factorize when a model or table needs it. Verify with str() or table() after recoding.