I use the read_sav
function from haven
package to import an SPSS file. Therefore I have column names and associate labels (class labelled
).
I lost the labels when I ordered the data frame. I can avoid the problem with a conversion in factor before order but is it a bug or a normal behavior ?
Here is a simple example.
DataForExample <- structure(list(CollectorNm = structure(c("Email Invitation 8",
"Email Invitation 8", "Email Invitation 8", "Email Invitation 8",
"Email Invitation 8", "Email Invitation 8"), label = "CollectorNm"),
q0001 = structure(c(1, 1, 1, 1, 1, 1), label = "Avez-vous déjà suivi la formation Atlas-Vente des 18 et 19 octobre ?", class = "labelled", labels = structure(c(1,
2), .Names = c("Oui, j'ai bien suivi cette formation.", "Non, je n'y ai pas participé."
))), q0002_0001 = structure(c(3, 3, 3, 2, 3, 3), label = "La formation dans son ensemble", class = "labelled", labels = structure(c(1,
2, 3, 4), .Names = c("pas du tout satisfait", "plutôt pas satisfait",
"plutôt satisfait", "très satisfait")))), .Names = c("CollectorNm",
"q0001", "q0002_0001"), class = c("tbl_df", "tbl", "data.frame"
), row.names = c(NA, -6L))
View(DataForExample) # OK
Toto <- DataForExample[order(DataForExample$q0001_0001),]
View(Toto) # NOK : the labels disappeared
Thanks
You need to load package with support for subsetting operations for class labelled
. It is better to load it after the haven
. There are at least two packages with such support: Hmisc
and expss
. Disclaimer: I am an author of the expss package.
UPDATE. Fix labelled class.
It seems that haven
doesn't set labelled
class for all variables with labels. So we need to fix it:
library(expss)
for(each in colnames(DataForExample)){
if(!("labelled" %in% class(DataForExample[[each]])) &&
(!is.null(var_lab(DataForExample[[each]])) ||
!is.null(val_lab(DataForExample[[each]]))
)) {
class(DataForExample[[each]]) = union("labelled", class(DataForExample[[each]]))
}
}
View(DataForExample) # OK
Toto <- DataForExample[order(DataForExample$q0001),]
View(Toto) # OK