Establish covariance matrix in R

Question

I am using two for loops to establish a covariance matrix in R. I have 2 data vectors and a covariance function f(x) between every 2 data points. My code is like this:

r12=function(x1,x2,H){
  ra12= matrix(0,length(x1),length(x2))
  for (i in 1:nrow(ra12)) 
  {
    for (j in 1:ncol(ra12)) 
    {
      raed12[i,j]=fx(x1[i]-x2[j],H)
    }
  }
  return(raed12)
}

How can I make this code faster?

For starters, how about changing your inner for loop so as to only compute the lower or upper triangle of the covariance matrix? — Mark L. Stone, Feb 24 '16 at 01:29
Is `fx` vectorized? If so, your code looks like longhand for computing an outer product. See the help for `outer`. — whuber, Feb 24 '16 at 01:30
@MarkL.Stone I had originally posed that as an answer, but I think that there are two different vectors, i.e. it may be a rectangular Gram matrix when x1 != x2. — Sycorax, Feb 24 '16 at 04:00

score 1 · Accepted Answer · edited May 23 '17 at 11:58

1

Consider an sapply() solution, passing in multiple arguments assuming fx() is vectorized:

r12 <- function(x1,x2,H){
          return(fx(x1-x2,H))
       }
mat <- sapply(vector1, r12, vector2, H)

edited May 23 '17 at 11:58

Community

1
1

answered Feb 24 '16 at 02:15

Parfait

104,375
17
94
125

I really thank you a lot, you have saved me a lot of time really – raK1 Feb 25 '16 at 02:52
It seems I do not need to Vectorize the function , also when I vectorize it is gives an error , can you explain that ? – raK1 Feb 25 '16 at 02:53

Establish covariance matrix in R

1 Answers1