How to convert for loop to apply in r [closed] - r

Question

Closed. This question needs debugging details. It is not currently accepting answers.
Edit the question to include desired behavior, a specific problem or error, and the shortest code necessary to reproduce the problem. This will help others answer the question.
Closed 3 years ago.
Improve this question
I have the following loop. I am trying to convert it into using the apply function instead of a loop, but I don't know how to rewrite the code.
for (i in 1:dim(Y)[2]) {
K = K_origin
print(i)
e = lmmlite::eigen_rotation(K, t(Y)[i,], use_cpp = T)
VC = lmmlite::fitLMM(
e$Kva,
e$y,
e$X,
reml = T,
use_cpp = T,
tol = 1e-6,
check_boundary = T
)
write.table(
VC$sigmasq_g,
"Vg_temp.txt",
row.names = F,
col.names = F,
append = T,
quote = F,
sep = "\n"
)
write.table(
VC$sigmasq_e,
"Ve_temp.txt",
row.names = F,
col.names = F,
append = T,
quote = F,
sep = "\n"
)
}
I want results like these
Vg Ve
1.15521325512487 0.755118863386436
0.579039221720728 1.21733212837417
0.372439354137817 0.296327744338075
0.0668396114713355 0.300417453013007
0.00771158861391208 0.100176380868691
0.210174870097273 0.141907482831872

Answer 1

R's apply functions has to be formulated as 1) something to iterate over, and 2) a function to apply to each element in (1).
But! Whether you will gain anything from converting your particular loop into an apply, is doubtful. If your loop is slow, I am guessing it is due to the operations performed, not that "R is slow on loops". If you only have 6 rows in Y, you will gain nothing from re-formulating the loop into an apply!
For your loop, each i is independent (as opposed to looping over i, when a result depends on the calculation on i-1). So that makes it very easy to re-formulate. Generally,
for (i in some_sequence) {
do something with i
}
can be reformulated to
my_do_something <- function(i) {
do something
}
for (i in some_sequence) {
my_do_something(i)
}
which can be again be directly reformulated to
sapply(some_sequence, my_do_something)
In your case, this would be
my_rotate <- function(i) {
e = lmmlite::eigen_rotation(K, t(Y)[i,], use_cpp = T)
VC = lmmlite::fitLMM( ... )
write.table(...)
write.table(...)
NULL
}
sapply(seq_len(dim(Y)[2]), my_rotate)
Notice how I added an NULL at the bottom of the function? That would be because apply will gather the returned values from the iterated function; write.table returns the written data invisible. Try the function without the last NULL and see what apply returns.
But wait, there's more!
Since you are iterating over particular rows (and asking about apply in particular), let's just drop the i stuff and feed the function the row:
my_rotate_row <- function(x) {
# you might or might not need to either use x as is, transpose it as t(x) or double transpose it, t(t(x)), to get the correct orientation.
# x is most likely an atomic vector, whereas `eigen_rotation` might be requiring either a row-vector or a column vector.
e = lmmlite::eigen_rotation(K, x, use_cpp = T)
VC = lmmlite::fitLMM( ... )
# lets collect data directly into a data.frame or matrix, instead of using files:
c(VC$sigmasq_g, VC$sigmasq_e)
}
Now you can use apply:
apply(Y, 2, my_rotate_row)