R three dimensional arrays

Question

I have an 3D array - arr

dim(arr)
[1] 2000   22    3

Let's call 2000 dim1, 22 dim2, and 3 dim3. I would like to do something like this:

for(d1 in 1:dim1){
    for(d2 in 1:dim2){
        #compute the mean across the 2000 values of arr[d1,d2]
        m<-mean(arr[d1,d2])
    }
}

Any idea?

What are you asking? How to make same computations faster or what? — Jouni Helske, Feb 27 '13 at 21:04
@Hemmo: he clearly would get errors with the code as it stands. Make the example smaller and then use it for testing. Any reference to the `arr`-object needs to have two commas inside the "[" call. — IRTFM, Feb 27 '13 at 21:05
(-1), after looking at the post and the OP's comments, I really can't figure out what the question is. — QkuCeHBH, Feb 27 '13 at 23:56

score 7 · Accepted Answer · answered Feb 27 '13 at 21:30

7

So the way your question is posed is a bit sloppy, but an example of what you might be trying to do is to take the average of each 2000 x 22 array for each of the 3 of them. here is how this would be done:

arr = array(1, dim=c(2000,22,3))
dim(arr)

m = NULL
dim3 = 3
for(d3 in 1:dim3){
    m[d3] = mean(arr[,,d3]) 
}

answered Feb 27 '13 at 21:30

Thanks. Let me try to further clarify it. Is there a way to convert this array into a list of 3 [22 X 2000] matrices? Then I can loop over the list and compute the mean for each matrix row. Thanks – user1701545 Feb 27 '13 at 21:52

IRTFM · Answer 2 · 2013-02-27T21:39:21.287

4

I'm guessing your might want this result:

m <- apply(arr, 1:2, mean)

Your current code would overwrite m at every inner iteration and you would end up with a single value, and it would also throw a dimension mismatch error when it encountered arr[d1,d2]. If you wanted to use a for-loop starategy, you would needed to define m as a dim1 x dim2 matrix and then populate its entries using m[d1,d2] <- mean(arr[d1, d2, ]).

Next time if you want to avoid all of those downvotes, why not show some testing with small example object:

arr <- array(1:5*4*3, c(5,4,3))

edited Feb 27 '13 at 21:39

answered Feb 27 '13 at 21:21

IRTFM

258,963
21
364
487

In the following Q&A you find little more code for a comparable situation and answer. It might be helpful to look at that code together with this answer: http://stackoverflow.com/questions/13475039/how-to-optimize-the-following-code-with-nested-while-loop-multicore-an-option – Jochem Feb 27 '13 at 21:27
My apologies for not being clear. Basically, not having much experience with R 3D arrays, my question is how do I perform an action like computing the mean over all d3 values, for each d2 dimension of each d1 dimension of an array with these dimensions: dim(arr) [1] d3 d2 d1 – user1701545 Feb 27 '13 at 21:35

score 1 · Answer 3 · answered Feb 27 '13 at 22:21

1

Never mind, I got it. That's how I would create a list of 3 [4 x 6] matrices:

arr = array(1, dim=c(6,4,3));
l<-list(length=dim(arr)[3]);
for (i in 1:dim(arr)[3]){
    l[[i]]<-arr[,,i];
}

cheers

answered Feb 27 '13 at 22:21

user1701545

5,706
14
49
80

R three dimensional arrays

3 Answers3