R：在R中的数据帧上使用cut，每列具有不同的断点

Question

I am trying to factorize a data frame, where the cut-offs would be min, median, max of each variable (column).

I have managed to do so, by creating a data frame "cuts" in which the respective values are stored, and using a for loop afterwards. However, I feel like it could be done more elegantly. Any idea would be welcome!

A reproducible example follows:

# Sample data frame
mydf <- na.omit(airquality)[1:20,1:4]

# Break points
cuts<-rbind(sapply(mydf,min),sapply(mydf,median),sapply(mydf,max))

# Data frame to keep factors
mydf.bin <- mydf

for (i in 1:ncol(mydf)) {
  mydf.bin[,i]<-cut(mydf[,i],cuts[,i],include.lowest=T)
}

mydf.bin

#I am looking for something like the following, except each column should have different break points
mybindf<-sapply(mydf, cut, c(0,50,350), include.lowest=T)

Answer 1

Why not just use an anonymous function?

mybindf <- sapply(mydf, function(x) {
    cuts <- c(min(x), median(x), max(x))
    cut(x, cuts, include.lowest = TRUE)
})

Better practice in this case might be to define the function separately, which makes for easier debugging and more readable code:

cut_min_med_max <- function(x) {
    cuts <- c(min(x), median(x), max(x))
    cut(x, cuts, include.lowest = TRUE)
}
mybindf <- sapply(mydf, cut_min_med_max)

The only difference between these solutions and your solution is that you generate the cut points separately from making the cuts, while here everything happens at once.

And for completeness, your original code can be vectorized:

mybindf <- as.data.frame(
    mapply(cut, mydf, cuts, MoreArgs = list(include.lowest = TRUE))
)

although you could just as easily have dropped both steps into the for loop.

Answer 2

You can use the bin function from the OneR package :

library(OneR)

# Sample data frame
mydf <- na.omit(airquality)[1:20,1:4]

# bin function is an enhanced version of cut for data frames
mydf.bin <- bin(mydf, nbins = 2, method = "content")

mydf.bin
##        Ozone    Solar.R        Wind      Temp
## 1    (15,41]  (170,334] (7.39,11.5] (64.5,74]
## 2    (15,41] (7.67,170] (7.39,11.5] (64.5,74]
## 3  (0.96,15] (7.67,170] (11.5,20.1] (64.5,74]
## 4    (15,41]  (170,334] (7.39,11.5] (57,64.5]
## 7    (15,41]  (170,334] (7.39,11.5] (64.5,74]
## 8    (15,41] (7.67,170] (11.5,20.1] (57,64.5]
## 9  (0.96,15] (7.67,170] (11.5,20.1] (57,64.5]
## 12   (15,41]  (170,334] (7.39,11.5] (64.5,74]
## 13 (0.96,15]  (170,334] (7.39,11.5] (64.5,74]
## 14 (0.96,15]  (170,334] (7.39,11.5] (64.5,74]
## 15   (15,41] (7.67,170] (11.5,20.1] (57,64.5]
## 16 (0.96,15]  (170,334] (7.39,11.5] (57,64.5]
## 17   (15,41]  (170,334] (11.5,20.1] (64.5,74]
## 18 (0.96,15] (7.67,170] (11.5,20.1] (57,64.5]
## 19   (15,41]  (170,334] (7.39,11.5] (64.5,74]
## 20 (0.96,15] (7.67,170] (7.39,11.5] (57,64.5]
## 21 (0.96,15] (7.67,170] (7.39,11.5] (57,64.5]
## 22 (0.96,15]  (170,334] (11.5,20.1] (64.5,74]
## 23 (0.96,15] (7.67,170] (7.39,11.5] (57,64.5]
## 24   (15,41] (7.67,170] (11.5,20.1] (57,64.5]

NB: the outer limits are moved away by 0.1% of the range to ensure that the extreme values both fall within the break intervals as is standard behaviour of the cut function.

(Full discosure: I am the author of this package)