使用sapply和tapply

Question

我努力纠正我的代码。给定此数据帧

S <- data.frame(Z1=c("A","A","A","D","D","A","A","A"),
                Z2=c("A","A","A","D","D","C","C","D"),
                K1=c(24,36,44,63,34,26,19,23),
                K2=c(12,24,13,16,23,25,12,34))

我应用了这个转换：

B <- sapply(1:2, function(x) {
    x1 <- S[c(x, x+2)]
    tapply(x1[,2], x1[,1], FUN=function(S) ceiling(median(S)))
})
colnames(B) <- c("G1","G2")

我希望将B设置为

       G1   G2
C       0   19
B       0    0
A      25   13
D      49   23

但我得到的却是这个错误：

Error in `colnames<-`(`*tmp*`, value = c("G1", "G2")) : 
  attempt to set 'colnames' on an object with less than two dimensions

Answer 1

你的问题之一是，R不知道你认为Z1和Z2是可以接受A，B，C，D值的分类变量。你告诉它的方式是factor类型。

S <- data.frame(Z1=c("A","A","A","D","D","A","A","A"),
                Z2=c("A","A","A","D","D","C","C","D"),
                K1=c(24,36,44,63,34,26,19,23),
                K2=c(12,24,13,16,23,25,12,34))

S$Z1 <- factor(S$Z1, levels=c("A", "B", "C", "D"))
S$Z2 <- factor(S$Z2, levels=c("A", "B", "C", "D"))

请注意，我必须明确地说明，A、B、C、D中的所有四个都是可能的，即使它们并不都出现。完成此操作后，您的转换函数将生成一个2D矩阵，colnames可应用于该矩阵。

B <- sapply(1:2, function(x) {
    x1 <- S[c(x, x+2)]
    tapply(x1[,2], x1[,1], FUN=function(S) ceiling(median(S)))
})
colnames(B) <- c("G1","G2")

然而，你没有在你期望的地方得到0，你得到的是NAs：

> B
  G1 G2
A 25 13
B NA NA
C NA 19
D 49 23

这是因为空集的中位数是未定义的。你可以用is.na来掩盖这一点

> B[is.na(B)] <- 0
> B
  G1 G2
A 25 13
B  0  0
C  0 19
D 49 23

此外，S[c(x, x+2)]的东西非常脆弱，如果我是您，我就不会在生产代码中依赖它。同样，在使用sapply(1:2, function(x) ...)时，函数对全局变量进行操作。

您可能会发现reshape2包更容易说服它去做您想做的事情。