R for modellers - Vignette 06

apply, lapply and friends

Julien Arino

Department of Mathematics

University of Manitoba*

* The University of Manitoba campuses are located on original lands of Anishinaabeg, Cree, Oji-Cree, Dakota and Dene peoples, and on the homeland of the Métis Nation.

Outline

• lapply
• “Advanced” lapply
• Avoid parameter variation loops with expand.grid

lapply

Very useful function (a few others in the same spirit: sapply, vapply, mapply)

Applies a function to each entry in a list/vector/matrix

Because there is a parallel version (parLapply) that we will see later, worth learning

l = list()
for (i in 1:10) {
        l[[i]] = runif(i)
}
lapply(X = l, FUN = mean)

or, to make a vector

unlist(lapply(X = l, FUN = mean))

or

sapply(X = l, FUN = mean)

“Advanced” lapply

Can “pick up” nontrivial list entries

l = list()
for (i in 1:10) {
        l[[i]] = list()
        l[[i]]$a = runif(i)
        l[[i]]$b = runif(2*i)
}
sapply(X = l, FUN = function(x) length(x$b))

gives

[1]  2  4  6  8 10 12 14 16 18 20

Just recall: the argument to the function you define is a list entry (l[[1]], l[[2]], etc., here)

Avoid parameter variation loops with expand.grid

# Suppose we want to vary 3 parameters
variations = list(
    p1 = seq(1, 10, length.out = 10),
    p2 = seq(0, 1, length.out = 10),
    p3 = seq(-1, 1, length.out = 10)
)

# Create the list
tmp = expand.grid(variations)
PARAMS = list()
for (i in 1:dim(tmp)[1]) {
    PARAMS[[i]] = list()
    for (k in 1:length(variations)) {
        PARAMS[[i]][[names(variations)[k]]] = tmp[i, k]     
    }
}

There is still a loop, but you can split this list, use it on different machines, etc. And can use parLapply