Variable Labels in Factorize and Characterize [Suggested Code Change]

[davidaarmstrong]

Please specify whether your issue is about:

• a suggested code or documentation change, improvement to the code, or feature request

Would it be possible to change the characterize and factorize functions as follows? This will propagate the variable label through to the newly created variable (if it exists).

I also added a couple of lines to factorize that would omit missing values in the label table that identify original responses that were coded as missing.

characterize.default <- function(x, ...) {
    xlab <- NULL
    if(!is.null(attributes(x)$label)){
        xlab <- attributes(x)$label
    }
    
    if (!is.null(attributes(x)$labels)) {
        x <- as.character(factorize(x, ...))
        if(!is.null(xlab)){
            attr(x, "label") <- xlab
        }
        x
    } else {
        x
    }
}

and

factorize.default <- function(x, ...) {
    xlab <- NULL
    if(!is.null(attributes(x)$label)){
        xlab <- attributes(x)$label
    }
    if (!is.null(attributes(x)$labels)) {
        labs <- na.omit(attributes(x)$labels)
        x <- factor(x, labs, names(labs), ...)
        if(!is.null(xlab)){
            attr(x, "label") <- xlab
        }
        x
    } else {
        x
    }
    
}```

[leeper]

Can you show me a before and after on what the difference is in output?

[davidaarmstrong]

Here is the output from factorize in the current version: ```R

library(rio) dat <- import("https://quantoid.net/files/rbe/r_example.dta") fdat <- factorize(dat) str(dat)

'data.frame': 10 obs. of 3 variables: $ x1: num 1 2 3 4 3 4 1 2 5 6 ..- attr(*, "label")= chr "First variable" ..- attr(*, "format.stata")= chr "%8.0g" $ x2: num 0 0 1 0 0 0 1 1 1 0 ..- attr(*, "label")= chr "Second variable" ..- attr(*, "format.stata")= chr "%8.0g" ..- attr(*, "labels")= Named num 0 1 .. ..- attr(*, "names")= chr "none" "some" $ x3: chr "yes" "no" "no" "yes" ... ..- attr(*, "label")= chr "Third variable" ..- attr(*, "format.stata")= chr "%3s"

str(fdat)

'data.frame': 10 obs. of 3 variables: $ x1: num 1 2 3 4 3 4 1 2 5 6 ..- attr(*, "label")= chr "First variable" ..- attr(*, "format.stata")= chr "%8.0g" $ x2: Factor w/ 2 levels "none","some": 1 1 2 1 1 1 2 2 2 1 $ x3: chr "yes" "no" "no" "yes" ... ..- attr(*, "label")= chr "Third variable" ..- attr(*, "format.stata")= chr "%3s"``` ``` Note that the attributes, particularly the variable label on x2 (in this case “Second Variable”) is not attached to x2 in the fdat data frame. With the proposed change, you would get the following: ```R

dat <- import("https://quantoid.net/files/rbe/r_example.dta") fdat <- factorize(dat) str(dat)

'data.frame': 10 obs. of 3 variables: $ x1: num 1 2 3 4 3 4 1 2 5 6 ..- attr(*, "label")= chr "First variable" ..- attr(*, "format.stata")= chr "%8.0g" $ x2: num 0 0 1 0 0 0 1 1 1 0 ..- attr(*, "label")= chr "Second variable" ..- attr(*, "format.stata")= chr "%8.0g" ..- attr(*, "labels")= Named num 0 1 .. ..- attr(*, "names")= chr "none" "some" $ x3: chr "yes" "no" "no" "yes" ... ..- attr(*, "label")= chr "Third variable" ..- attr(*, "format.stata")= chr "%3s"

str(fdat)

'data.frame': 10 obs. of 3 variables: $ x1: num 1 2 3 4 3 4 1 2 5 6 ..- attr(*, "label")= chr "First variable" ..- attr(*, "format.stata")= chr "%8.0g" $ x2: Factor w/ 2 levels "none","some": 1 1 2 1 1 1 2 2 2 1 ..- attr(*, "label")= chr "Second variable" $ x3: chr "yes" "no" "no" "yes" ... ..- attr(*, "label")= chr "Third variable" ..- attr(*, "format.stata")= chr "%3s" ``` Note that the variable label now follows x2 into the factorized data frame. This is particularly useful when you will eventually want to export the variable back to something like Stata or SPSS. In this case, the current version of the function would only return the variable labels for the non factors. The proposed change would return variable labels for all variables. Best, Dave.

[leeper]

Got it. Thanks. Yes, this is a bug - appreciate the suggested fix!