CODEWITHSUNDEEP
arraysrmatrix
Nemesi
Nemesi

Reputation: 801

R 3d array to 2d matrix

Let's assume I have a 3d array of dimensions (x,y,z) and would like to restructure my data as a matrix of dimensions (x*y,z), something like:

my_array <- array(1:600, dim=c(10,5,12))
my_matrix<-data.frame()

for (j in 1:5) {
  for (i in 1:10) {
     my_matrix <- rbind (my_matrix, my_array[i,j,1:12])
 }
}

Could you suggest a faster and more elegant way?

thanks

Upvotes: 12

Views: 16001

Answers (3)

Good Fit
Good Fit

Reputation: 1316

Both @akrun and @Lars Arne Jordanger's solutions work and generate the same results.

The two solutions work by:

(1) concatenating the first rows of all matrices together and placing these rows in the top of the combined matrix; and then

(2) concatenating the second rows of all matrices together and placing these rows under the concatenation of the first rows, and so on.

The following example illustrates the idea very well:

> threeDimArray <- array( NA, dim=c(3,3,4) )
> dims <- dim( threeDimArray )
> 
> constants <- c(1, 10, 100)
> for( id in 1:length(constants) ){
   const <- constants[id]
   threeDimArray[id,,] <- matrix( (1:prod(dims[2:3]))*const, dims[2], dims[3] )
 }
> threeDimArray[1,,]
     [,1] [,2] [,3] [,4]
[1,]    1    4    7   10
[2,]    2    5    8   11
[3,]    3    6    9   12
> threeDimArray[2,,]
     [,1] [,2] [,3] [,4]
[1,]   10   40   70  100
[2,]   20   50   80  110
[3,]   30   60   90  120
> threeDimArray[3,,]
     [,1] [,2] [,3] [,4]
[1,]  100  400  700 1000
[2,]  200  500  800 1100
[3,]  300  600  900 1200
> # solution 1:
> twoDimMat <- matrix(threeDimArray, prod(dims[1:2]), dims[3])
> twoDimMat
      [,1] [,2] [,3] [,4]
 [1,]    1    4    7   10
 [2,]   10   40   70  100
 [3,]  100  400  700 1000
 [4,]    2    5    8   11
 [5,]   20   50   80  110
 [6,]  200  500  800 1100
 [7,]    3    6    9   12
 [8,]   30   60   90  120
 [9,]  300  600  900 1200
> 
> # solution 2: 
> threeDArray <- threeDimArray
> dim(threeDArray) <- c(prod( dims[1:2] ), dims[3])
> threeDArray
      [,1] [,2] [,3] [,4]
 [1,]    1    4    7   10
 [2,]   10   40   70  100
 [3,]  100  400  700 1000
 [4,]    2    5    8   11
 [5,]   20   50   80  110
 [6,]  200  500  800 1100
 [7,]    3    6    9   12
 [8,]   30   60   90  120
 [9,]  300  600  900 1200
>

Upvotes: 1

akrun
akrun

Reputation: 887901

We can convert to a matrix by calling the matrix and specifying the dimensions

res <- matrix(my_array, prod(dim(my_array)[1:2]), dim(my_array)[3])
all.equal(as.matrix(my_matrix), res, check.attributes=FALSE)
#[1] TRUE

NOTE: This will not change the original 'my_array`. Also, in fact, the code can be simplified to 

matrix(my_array, 10*5, 12)

and make it compact. 

nchar("matrix(my_array, 10*5, 12)")
#[1] 26

nchar("dim(my_array) <- c(10 * 5 , 12)")
#[1] 31

Upvotes: 11

Lars Arne Jordanger
Lars Arne Jordanger

Reputation: 671

Change the dimension of the array:

dim(my_array) <- c(10 * 5 , 12)

Upvotes: 13

Related Questions