Replacing pair of element of symmetric matrix with NA-CodePudding

I have a positive definite symmetric matrix. Pasting the matrix generated using the following code:

set.seed(123)
m <- genPositiveDefMat(
    dim = 3, 
    covMethod = "unifcorrmat",
    rangeVar = c(0,1) )

x <- as.matrix(m$Sigma)
  diag(x) <- 1

x

#Output
          [,1]       [,2]       [,3]
[1,]  1.0000000 -0.2432303 -0.4110525
[2,] -0.2432303  1.0000000 -0.1046602
[3,] -0.4110525 -0.1046602  1.0000000

Now, I want to run the matrix through iterations and in each iteration I want to replace the symmetric pair with NA. For example,

Iteration 1: x[1,2] = x[2,1] <- NA

Iteration2: x[1,3] = x[3,1] <- NA

and so on....

My idea was to check using a for loop

Prototype:

for( r in 1:nrow(x)
for( c in 1:ncol(x)

if x[r,c]=x[c,r]<-NA

else

x[r,c]

The issue with my code is for row 1 and column 1, the values are equal hence it sets to 0 (which is wrong). Also, the moment it is not NA it comes out of the loop.

Appreciate any help here.

Thanks

CodePudding user response：

If you need the replacement done iteratively, you can use the indexes of values represented by upper.tri(x)/lower.tri to do the replacements pair-by-pair. That will allow you to pass the results to a function before/after each replacement, e.g.:

sel <- sapply(c(lower.tri, upper.tri), \(f) which(f(mat)) )
#     [,1] [,2]
#[1,]    2    4 ##each row represents the lower/upper pair
#[2,]    3    7
#[3,]    6    8

for( i in seq_len(nrow(sel)) ) {
    mat[ sel[i,] ] <- NA
    print(mean(mat, na.rm=TRUE))
}
#[1] 0.2812249
#[1] 0.5581359
#[1] 1