在R中剪切函数以考虑重复出现的数据

Question

Say I have a range of data indicating the ages of individuals in years. Such that

ages <- sample(40:80, 30, replace = F)

Now I want to plot (boxplot) against another variable ~ say weight

But I want to cut the ages samples into the following catergories <50, >50, >60. >70. So that an individuals weight, who is 66, will be used for both the >50 and >60 plots

My understanding is I use the cut command

age.category <- cut(ages, breaks = c(40, 50, 60, 70, 80) ........)

But how do I format to accounnt for repeated data, when I want

labels = c("x < 50", "x > 50", "x > 60", "x > 70")

Answer 1

Here's an alternative "multicut" function which would allow you to specify arbitrary breaks. This will repeat values, once for each group that value appears. So if you had "64" it would output both 64, "x>50" and 64, "x>60" .

#sample data
set.seed(15)
ages <- sample(40:80, 30, replace = F)
weights <- 100 + ages*.2 + rnorm(30, 0 , 20)

#custom breaks
#named list, names-categories, values = 2-vector with min/max
breaks<-list(
   "x<50" =c(-Inf, 50),
   "x>50" =c(50, Inf),
   "x>60" = c(60, Inf),
   "x>70" = c(70, Inf)
)

Now we define the main helper function multicut

multicut <- function(x, breaks, vals=x, left.closed=TRUE, right.closed=FALSE, 
  x.name=if(missing(vals)) deparse(substitute(x)) else deparse(substitute(vals)),
  group.name="group") {

    unrowname <- function(x) {rownames(x)<-NULL; x}
    if (is.data.frame(vals)) {
        if(missing(x.name)) x.name<-names(vals)
        vals = Map(unrowname, split(vals, 1:nrow(vals)))
    }
    stopifnot(length(vals) == length(x))
    grp <- lapply(x, function(x) {
        mapply(function(z, br,l,r) {
            left<-if (l) z>=br[1] else z>br[1]
            right<-if (r) z<=br[2] else z<br[2]
            left & right
        }, x, breaks, left.closed, right.closed)
    })
    df <- do.call(rbind.data.frame, 
        Map(cbind.data.frame,  
        g=lapply(grp, function(z) if(any(z)) names(breaks)[z] else NA),
        x=vals))
    df[[1]] <- factor(df[[1]], levels=names(breaks))
    names(df) <- c(group.name, x.name)
    df
}

Now we use it on the sample data

dd <- multicut(ages, breaks, weights)
boxplot(weights~group, dd)

The three important parameters to multicut are x which contains the values you use to wish for categorization, breaks which is the named list of min/max values for each group, and optionally vals which is either a vector or data.frame you want to split based on x and breaks . Here we want to use age to split up the weights .