[R] Change values in a dateframe

[Berend Hasselman]

On 24-07-2013, at 08:39, Arnaud Michel <michel.arnaud at cirad.fr> wrote:

> Hello
> 
> I have the following problem :
> The dataframe TEST has multiple lines for a same person because :
> there are differents values of Nom or differents values of Prenom
> but the values of Matricule or Sexe or Date.de.naissance are the same.
> 
> TEST <- structure(list(Matricule = c(66L, 67L, 67L, 68L, 89L, 90L, 90L,
> 91L, 108L, 108L, 108L), Nom = structure(c(1L, 2L, 2L, 4L, 8L,
> 5L, 6L, 9L, 3L, 3L, 7L), .Label = c("CHICHE", "GEOF", "GUTIER",
> "JACQUE", "LANGUE", "LANGUE-LOPEZ", "RIVIER", "TRU", "VINCENT"
> ), class = "factor"), Prenom = structure(c(8L, 3L, 4L, 5L, 1L,
> 2L, 2L, 9L, 6L, 7L, 7L), .Label = c("Edgar", "Elodie", "Jeanine",
> "Jeannine", "Michel", "Michele", "Michèle", "Michelle", "Victor"
> ), class = "factor"), Sexe = structure(c(1L, 1L, 1L, 2L, 2L,
> 1L, 1L, 2L, 1L, 1L, 1L), .Label = c("Féminin", "Masculin"), class = "factor"),
>    Date.de.naissance = structure(c(4L, 2L, 2L, 7L, 6L, 5L, 5L,
>    1L, 3L, 3L, 3L), .Label = c("03/09/1940", "04/03/1946", "07/12/1947",
>    "18/11/1945", "27/09/1947", "29/12/1936", "30/03/1935"), class = "factor")), .Names = c("Matricule",
> "Nom", "Prenom", "Sexe", "Date.de.naissance"), class = "data.frame", row.names = c(NA,
> -11L))
> 
> 
> I would want to make homogeneous the information and would like built 2 dataframes :
> df1 wich has the value of Nom and Prenom of the first lines of TEST when there are different values. The other values (Matricule or Sexe or Date.de.naissance) are unchanged
> 
> df1 <- structure(list(Matricule = c(66L, 67L, 67L, 68L, 89L, 90L, 90L,
> 91L, 108L, 108L, 108L), Nom = structure(c(1L, 2L, 2L, 4L, 6L,
> 5L, 5L, 7L, 3L, 3L, 3L), .Label = c("CHICHE", "GEOF", "GUTIER",
> "JACQUE", "LANGUE", "TRU", "VINCENT"), class = "factor"), Prenom = structure(c(6L,
> 3L, 3L, 4L, 1L, 2L, 2L, 7L, 5L, 5L, 5L), .Label = c("Edgar",
> "Elodie", "Jeanine", "Michel", "Michele", "Michelle", "Victor"
> ), class = "factor"), Sexe = structure(c(1L, 1L, 1L, 2L, 2L,
> 1L, 1L, 2L, 1L, 1L, 1L), .Label = c("Féminin", "Masculin"), class = "factor"),
>    Date.de.naissance = structure(c(4L, 2L, 2L, 7L, 6L, 5L, 5L,
>    1L, 3L, 3L, 3L), .Label = c("03/09/1940", "04/03/1946", "07/12/1947",
>    "18/11/1945", "27/09/1947", "29/12/1936", "30/03/1935"), class = "factor")), .Names = c("Matricule",
> "Nom", "Prenom", "Sexe", "Date.de.naissance"), class = "data.frame", row.names = c(NA,
> -11L))
> 
> df2 wich has the value of Nom and Prenom of the last lines of TEST when there are different values. The other values (Matricule or Sexe or Date.de.naissance) are unchanged.
> 
> df2 <- structure(list(Matricule = c(66L, 67L, 67L, 68L, 89L, 90L, 90L,
> 91L, 108L, 108L, 108L), Nom = structure(c(1L, 2L, 2L, 3L, 6L,
> 4L, 4L, 7L, 5L, 5L, 5L), .Label = c("CHICHE", "GEOF", "JACQUE",
> "LANGUE-LOPEZ", "RIVIER", "TRU", "VINCENT"), class = "factor"),
>    Prenom = structure(c(6L, 3L, 3L, 4L, 1L, 2L, 2L, 7L, 5L,
>    5L, 5L), .Label = c("Edgar", "Elodie", "Jeannine", "Michel",
>    "Michèle", "Michelle", "Victor"), class = "factor"), Sexe = structure(c(1L,
>    1L, 1L, 2L, 2L, 1L, 1L, 2L, 1L, 1L, 1L), .Label = c("Féminin",
>    "Masculin"), class = "factor"), Date.de.naissance = structure(c(4L,
>    2L, 2L, 7L, 6L, 5L, 5L, 1L, 3L, 3L, 3L), .Label = c("03/09/1940",
>    "04/03/1946", "07/12/1947", "18/11/1945", "27/09/1947", "29/12/1936",
>    "30/03/1935"), class = "factor")), .Names = c("Matricule",
> "Nom", "Prenom", "Sexe", "Date.de.naissance"), class = "data.frame", row.names = c(NA,
> -11L))
> 

Something like this

r1 <- droplevels(do.call(rbind,lapply(split(TEST,TEST$Matricule),
                    FUN=function(x) {x[,c("Nom","Prenom")] <- x[1,c("Nom","Prenom"),drop=TRUE];x})))
rownames(r1) <- NULL 
r1

r2 <- droplevels(do.call(rbind,lapply(split(TEST,TEST$Matricule),
                    FUN=function(x) {x[,c("Nom","Prenom")] <- x[nrow(x),c("Nom","Prenom"),drop=TRUE];x})))
rownames(r2) <- NULL
r2

#> identical(r1,df1)
#[1] TRUE
#> identical(r2,df2)
#[1] TRUE

Note: I had to change the Prenom and Sexe columns because of encoding issues. but that shouldn't have any influence on the above.

Berend