apply() in R with user-defined function

问题:

I have a data frame with votes and party labels arranged thus

dat <- data.frame( v1=c(25, 0, 70), 
                   v2=c(75, 100, 20), 
                   v3=c(0, 0, 10), 
                   l1=c("pA", ".", "pB"), 
                   l2=c("pB", "pC", "pC"), 
                   l3=c(".", ".", "pD") )

so that each row is a unit of analysis. Only vote-getting parties need consideration and this function extracts positive votes or the corresponding labels

getpos <- function(vector, vorl="v"){ # change to "l" to report labels
    vot <- vector[grep( "v", colnames(vector) )]; 
    lab <- vector[grep( "l", colnames(vector) )];
    if (vorl=="v") {vot[vot>0]} else {lab[vot>0]};
}
getpos(dat[1,])           # votes for obs 1
getpos(dat[1,], vorl="l") # labels for obs 1

I wish to run function getpos in every row of data frame dat in order to produce lists with vote/label vectors of different length. Applying the function does not return what I expect:

apply(X=dat, MARGIN=1, FUN=getpos, vorl="l")

Can anyone spot the problem? And related, can this be achieved more efficiently?

回答1:

What's happening here is that the rows in the dataframe no longer have column names after being extracted by apply (but they do have names):

Try:

getpos <- function(x, vorl="v"){ 
     vot <- x[grep( "v", names(x) )] ;  lab <- x[grep( "l", names(x) )];
     if (vorl=="v") {vot[vot>0]} else {lab[vot>0]};
 }

> apply(dat, MARGIN=1, FUN=function(x2) getpos(x2, vorl="l") )
#-------------
[[1]]
  l1 
"pA" 

[[2]]
  l2 
"pC" 

[[3]]
  l1   l3 
"pB" "pD"