Joschi Joschi - 4 months ago 10
R Question

Simple z-transformation over all columns

I just wrote a very basic function for the z tranformation in a Matrix over all columns. It looks like this:

sapply(MyObject, function(x){(x-mean(x))/sd(x)})


I randomly checked my function for some cells within the Matrix and it seems to work fine. Still I wanted to confirm the function is okay, because I'm very new to R and I could not find any good examples on the internet.

Answer

You should look at scale() which does this for you.

Your function is close to being correct; you should add na.rm = TRUE to both the sd() and mean() function calls.

I would write (if not using scale()) the function using sweep() instead of the sapply(). E.g.

ztran <- function(x, na.rm = TRUE) {
    mns <- colMeans(x, na.rm = na.rm)
    sds <- apply(x, 2, sd, na.rm = na.rm)
    x <- sweep(x, 2, mns, "-")
    x <- sweep(x, 2, sds, "/")
    x
}

In use we have

> df <- data.frame(matrix(1:9, ncol = 3))
> ztran(df)
  X1 X2 X3
1 -1 -1 -1
2  0  0  0
3  1  1  1
> scale(df)
     X1 X2 X3
[1,] -1 -1 -1
[2,]  0  0  0
[3,]  1  1  1
attr(,"scaled:center")
X1 X2 X3 
 2  5  8 
attr(,"scaled:scale")
X1 X2 X3 
 1  1  1

sweep is a very useful vectorised tool for this sort of operation. Notice also that sapply() simplifies to a matrix, which may not be what you wanted. sweep() doesn't do this:

> class(ztran(df))
[1] "data.frame"
> class(sapply(df, function(x){(x-mean(x))/sd(x)}))
[1] "matrix"