CODEWITHSUNDEEP
rmatrixdplyrapply
sriramn
sriramn

Reputation: 2378

Rolling window pattern update in R

I would like to apply a rolling window pattern update on a matrix. Say I have a matrix of zeros and ones like so:

mat <- matrix(c(0,1,1,0,1,1,0,0,1,1,1,0,1,0,0,0),nrow=4)

0   1   1   1
1   1   1   0
1   0   1   0
0   0   0   0

I want to scan each row of the matrix 2 values at a time and create a matrix of labels based on this set of rules:

0   0   - then label as A
0   1   - then label as B
1   0   - then label as B
1   1   - then label as C

The desired output for the above matrix is: (the NA values appear because the first value does not have a value before it)

NA  B   C   C
NA  C   C   B
NA  B   B   B
NA  A   A   A

I am trying to implement this in R. Is there a way to do this using one of the apply methods? Or is there a quick approach storing this data as a data.frame and processing using packages such as dplyr or data.table? Currently, I am processing it as a for loop (which is, of course, very very slow).

Upvotes: 3

Views: 98

Answers (1)

Frank
Frank

Reputation: 66819

The rolling operation here acts on pairwise combinations, so you can just define the function:

f <- function(x,y) LETTERS[1:3][x+y+1L]

and then apply it to the appropriate pairs:

nc  <- ncol(m)
res <- matrix(,length(m)/nc,nc)

res[,-1] <- f(m[,-nc], m[,-1])

#      [,1] [,2] [,3] [,4]
# [1,] NA   "B"  "C"  "C" 
# [2,] NA   "C"  "C"  "B" 
# [3,] NA   "B"  "B"  "B" 
# [4,] NA   "A"  "A"  "A"

Upvotes: 6

Related Questions