CODEWITHSUNDEEP
rmatrixdataframe
MAOC
MAOC

Reputation: 635

Find element in matrix and substitute

I have data.frame of survey results. The responses are in Portuguese language and I need to substitute some responses, for instance, "Não sabe" and "Não respondeu" by "Ns/Nr". Each column of the matrix is of factor class. Can I generalize this procedure for any labels? I tried something like this

mydata[mydata %in% c("Não sabe", "Não respondeu")] <- "Ns/Nr"

But, it doesn't work. In fact, when I tried: 

mydata[mydata %in% c("Não sabe", "Não respondeu")]<- "Ns/Nr"

there is not error, but, when I do 

freq(mydata$Q_9)

there is no "Ns/Nr" label and the other frequencies remain the same. And, when I do 

mydata[mydata == "Não respondeu"]<- "Ns/Nr"

there are warnings like 

In `[<-.factor`(`*tmp*`, thisvar, value = "Ns/Nr") :
invalid factor level, NA generated

In this case, when I do freq(mydata$Q_9), the frequence of the "Não respondeu" is zero and NA takes the frequence of the old "Não respondeu".

Upvotes: 1

Views: 55

Answers (1)

akrun
akrun

Reputation: 887291

You could convert the 'factor' class to 'character' before substituting i.e.

 mydata[] <- lapply(mydata, as.character)
 mydata[] <- lapply(mydata, function(x) {
        x[ x %in%c("Não sabe", "Não respondeu")] <- 'Ns/Nr'
         x})

Or without converting to 'character' class we can use recode from car. The advantage is that you can have new levels 'Ns/Nr' updated in the 'factor' column while dropping the levels replaced.

 library(car)
 mydata[] <- lapply(mydata, function(x) 
         recode(x, "c('Não sabe', 'Não respondeu')='Ns/Nr'") )

Upvotes: 1

Related Questions