将函数应用于利用列方式的数据框（~ Pareto scaling）

Question

在数据框中，我想将每个值除以列值的标准差的平方根（~ Pareto Scaling）。我已从现有包中获取代码 (https://github.com/cran/RFmarkerDetector/blob/master/R/scaling.R)

paretoscale <- function(data) {
    # Here we perform centering
    x.centered <- apply(x, 2, function(x) x - mean(x))
    # Then we perform scaling on the mean-centered matrix
    x.sc <- apply(x.centered, 2, function(x) x/sqrt(sd(x)))
    x.sc <- cbind(sample_classes, x.sc)

x.centered <- apply(x, 2, function(x) x - mean(x) 是不是应该像 x - mean(column where x is) 那样做？你能解释一下它是如何工作的吗？

Answer 1

变量名称可能有点缺乏教学意义，尤其是对于新手而言。让我们重写apply 部分，以免混淆读者。

paretoscale <- function(data) {
    # Here we perform centering
    x.centered <- apply(x, 2, function(col) col - mean(col))
    # Then we perform scaling on the mean-centered matrix
    x.sc <- apply(x.centered, 2, function(col) col/sqrt(sd(col)))
    x.sc <- cbind(sample_classes, x.sc)

apply(x, 2, function(col) col - mean(col)) 的作用是在对象x（data.frame 或矩阵）上按列运行。对于每一列，它会找到它的平均值并减去每个元素的平均值。

下面是apply 与for 循环的比较。

xy <- data.frame(matrix(1:9, ncol = 3))

apply(X = xy, MARGIN = 2, FUN = function(col) col - mean(col))

     X1 X2 X3
[1,] -1 -1 -1
[2,]  0  0  0
[3,]  1  1  1

# Create an empty object
newxy <- xy
newxy[] <- NA

# Work column-wise
for (i in 1:ncol(xy)) {
  col <- xy[, i]
  # Calculate mean and substract it from all elements of the column
  newxy[, i] <- col - mean(col)
}
newxy

  X1 X2 X3
1 -1 -1 -1
2  0  0  0
3  1  1  1