target encoding gives error Error in l[is.na(l)] <- m : replacement has length zero-CodePudding

I have tried to make a reproducible example although I do not succeed.

I try to do target encoding with the function below from rbloggers:

function(x, y, sigma = NULL) {
  d <- aggregate(y, list(factor(x, exclude = NULL)), mean, na.rm = TRUE)
  m <- d[is.na(as.character(d[, 1])), 2]
  l <- d[, 2]
  names(l) <- d[, 1]
  l <- l[x]
  l[is.na(l)] <- m
  if (!is.null(sigma)) {
    l <- l * rnorm(length(l), mean = 1, sd = sigma)
  }
  l
}

It runs fine on the example data:

data1 <- data.frame(factor_var = c("a", "a", "a", "a", "b", "b"),
                    target = c(TRUE, FALSE, FALSE, TRUE, FALSE, FALSE))

which gives:

> l
  a   a   a   a   b   b 
0.5 0.5 0.5 0.5 0.0 0.0

Although when I apply it to my own dataset it gives the error below:

Error in l[is.na(l)] <- m : replacement has length zero

The error is thrown by l[is.na(l)]

m is " numeric (empty) " although this seems not the cause.

Thanks a lot!

my own data:

structure(c(2L, 2L, 2L, 2L, 2L, 2L, 2L, 3L, 3L, 3L), levels = c("Albania_Albanian Cup", 
"Albania_Superliga", "Argentina_Copa Argentina", "Argentina_Copa de la Superliga"), class = "factor")

CodePudding user response：

The issue seems to be that the factor data may have some levels for which the data is not present, thus

l <- l[x]
> l
Argentina_Copa Argentina Argentina_Copa Argentina Argentina_Copa Argentina Argentina_Copa Argentina Argentina_Copa Argentina Argentina_Copa Argentina 
               0.3333333                0.3333333                0.3333333                0.3333333                0.3333333                0.3333333 
Argentina_Copa Argentina                     <NA>                     <NA>                     <NA> 
               0.3333333                       NA                       NA                       NA

creates some NA columns for those unused levels. Instead, we may have to use droplevels to remove the unused levels

f2 <- function(x, y, sigma = NULL) {
  x <- droplevels(x)
  d <- aggregate(y, list(factor(x, exclude = NULL)), mean, na.rm = TRUE)
  m <- d[is.na(as.character(d[, 1])), 2]
  l <- d[, 2]
  names(l) <- d[, 1]
  l <- l[x]
  l[is.na(l)] <- m
  if (!is.null(sigma)) {
    l <- l * rnorm(length(l), mean = 1, sd = sigma)
  }
  l
}

-testing

> f2(x, y)
       Albania_Superliga        Albania_Superliga        Albania_Superliga        Albania_Superliga        Albania_Superliga        Albania_Superliga 
               0.7142857                0.7142857                0.7142857                0.7142857                0.7142857                0.7142857 
       Albania_Superliga Argentina_Copa Argentina Argentina_Copa Argentina Argentina_Copa Argentina 
               0.7142857                0.3333333                0.3333333                0.3333333 
# OP's original function

> f1(x, y)
Error in l[is.na(l)] <- m : replacement has length zero

OR another option is to match with character converted vector instead of factor

> l[as.character(x)]
       Albania_Superliga        Albania_Superliga        Albania_Superliga        Albania_Superliga        Albania_Superliga        Albania_Superliga 
               0.7142857                0.7142857                0.7142857                0.7142857                0.7142857                0.7142857 
       Albania_Superliga Argentina_Copa Argentina Argentina_Copa Argentina Argentina_Copa Argentina 
               0.7142857                0.3333333                0.3333333                0.3333333 
> l[x]
Argentina_Copa Argentina Argentina_Copa Argentina Argentina_Copa Argentina Argentina_Copa Argentina Argentina_Copa Argentina Argentina_Copa Argentina 
               0.3333333                0.3333333                0.3333333                0.3333333                0.3333333                0.3333333 
Argentina_Copa Argentina                     <NA>                     <NA>                     <NA> 
               0.3333333                       NA                       NA                       NA 

f3 <- function(x, y, sigma = NULL) {
  d <- aggregate(y, list(factor(x, exclude = NULL)), mean, na.rm = TRUE)
  m <- d[is.na(as.character(d[, 1])), 2]
  l <- d[, 2]
  names(l) <- d[, 1]
  l <- l[as.character(x)]
  l[is.na(l)] <- m
  if (!is.null(sigma)) {
    l <- l * rnorm(length(l), mean = 1, sd = sigma)
  }
  l
}