Apply Family

We saw how to parse structures of data with "for" loops. As a reminder, this is how you would print the column mean values of a matrix and of a list using "for":

> ## Creation of a matrix
> mat <- matrix(data=1:6, nrow=2, ncol=3, byrow=T)
>
> ## Display the column mean values
> for(i in seq_len(ncol(mat))){
+ cat(mean(mat[,i]), "\n")
+ }
>
> ## Creation of a list with two elements
> mylist <- list(c(23,24,56), c(89,45,65))
>
> ## Display the mean value of each vector of the list
> for(i in seq_len(length(mylist))){ 
+ cat(mean(mylist[[i]]), "\n")
+ }

If you remember, I warned you that using "for" is not a good idea because of its poor computational performances. Instead, the function apply enables to do the same thing. It takes three main arguments: the matrix, the MARGIN and the function. MARGIN should take the value 1 if the matrix should be read by row, or the value 2 by column. The function used below is the same as in the example above: mean. Enter ?apply in your console to read the documentation of this function.

> apply(mat, MARGIN = 2, mean)

That much more efficient isn't it? The same thing can be done on list with the function lapply (apply on list, l = list). Because a list does not have 2 dimensions, the option MARGIN is not used:

> result <- lapply(mylist, mean)

Now let see what is the type of the result returned by lapply:

> result
> is(result)
> length(result)

"result" is a list of two elements each containing a mean value. As you would want to perform further operations on this means, you would like to transform them to a numeric vector. As you saw before you could use unlist or as.numeric:

> unlist(result)
> as.numeric(result)

A third possibility is to use sapply which is doing the same thing than lapply but returns a vector:

> result <- sapply(mylist, mean)
> result
> is(result)

Finally, the last example of this section shows you how to use lapply/sapply on more than one structure. For that, you can use mapply where 'm' strands for 'multiple'. You understand now why mapply is the multiple version of lapply and sapply. It runs apply on multiple lists or vectors:

> ## Creation of two lists with two elements
> mylist1 <- list(c(23,24,56), c(89,45,65))
> mylist2 <- list(c(65,12,78), c(23,79,32))
>
> ## Display the mean value of each vector of the lists
> mapply(function(x, y){
cat("The mean value of ", x, " is ", mean(x), "\n")
cat("The mean value of ", y, " is ", mean(y), "\n")
}, mylist1, mylist2)

Exercise 22:

• Create two numeric vectors of equal size
• Use mapply to perform a 2 by 2 sum of values
• Could you use a more clever solution?