Randomly sampling and averaging data in a matrix

Question

I have a 200x200 matrix containing count information. I would like to randomly sample a 5 row x 5 column subset within my matrix and take the average of the counts within this subset. I want to do this 8 times randomly throughout my matrix (so maybe use a for-loop?) and then take the averages of these 8. I also need to use restrictions so that the random number generator doesn't sample at the very edge of my matrix.

I am new to R and programming in general, so a push in the right direction would be very helpful.

Is this sampling with replacements? Are the subsets allowed to overlap. — Sven Hohenstein
– Sven Hohenstein, Commented Jan 27, 2014 at 21:25
This is not sampling with replacements. Yes the subsets are allowed to overlap. I want to randomly sample a 5x5 portion of my matrix and find the average of that 5x5 portion. I want to do this 8 times and then take the average of the averages from the 5x5 portions. Sorry if this is confusing. And thank you for your help! — Mary R
– Mary R, Commented Jan 29, 2014 at 21:34

Sven Hohenstein · Accepted Answer · 2014-01-29 22:40:17Z

1

Here's an approach:

# an example 10x10 matrix
mat <- matrix(1:100, 10)
nc <- ncol(mat) # number of columns
nr <- nrow(mat) # number of rows

size <- 5 # size of the subset matrix
nmat <- 8 # number of submatrices

# sample indices of submatrices
set.seed(1)
idxc <- sample(seq(2, nc - size), size = nmat, replace = TRUE)
idxr <- sample(seq(2, nr - size), size = nmat, replace = TRUE)

# create a list of 8 submatrices
res <- mapply(function(x, y) mat[seq(x, x + size - 1), seq(y, y + size - 1)],
              idxr, idxc, SIMPLIFY = FALSE)

# calculate the average of the averages
mean(unlist(res))
# [1] 53.875

edited Jan 29, 2014 at 22:40

answered Jan 27, 2014 at 21:32

Sven Hohenstein

82k17 gold badges150 silver badges173 bronze badges

Sign up to request clarification or add additional context in comments.

3 Comments

Mary R Over a year ago

The first part works great! Thank you! For the averaging though, I would like to find the average of the counts in the 5x5 subset (1 number). I would like to do this for all 8 different subsets. I would then like to find the average of these 8 subsets (instead of finding the average of the different cell values). This will give me 1 number at the end.

Mary R Over a year ago

Is there a way to get the code to randomly sample different parts of the matrix every time I run it? Currently, it gives me the same answer each time I run it.

Sven Hohenstein Over a year ago

@MaryR Do not run set.seed(1). It resets the random number generator to a certain state. It is not necessary for the approach. I used it for reproducibility.

Brandon Bertelsen · Accepted Answer · 2014-01-28 01:00:12Z

0

An alternative that assumes you're always pulling out a square. (ie, the rows/columns are right beside each other).

y <- matrix(round(rep(runif(200),200)*100),ncol=200,nrow=200)

sq <- seq(1,196,1) 
blocks <- sample(sq,8) # no replacements of squares

mean(
  sapply(blocks, 
    function(x) mean(y[x:(4+x),x:(4+x)])
  )
)

edited Jan 28, 2014 at 1:00

answered Jan 28, 2014 at 0:42

Brandon Bertelsen

44.8k37 gold badges170 silver badges262 bronze badges

Collectives™ on Stack Overflow

Randomly sampling and averaging data in a matrix

2 Answers 2

3 Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

3 Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related