I there a R function for finding a minimum between to pairs?

Question

I need to identify pairs of variables that are dominated: if both values of the pair are lower than the other pairs in the data.

I already tried the functions min or pmin but I am not sure if they are the most appropriate.

   a = matrix(c(50,70), ncol = 2)
   b = matrix(c(45,85), ncol = 2)

  df = rbind(a,b)

Dominance <- function(a){
   for (i in 1:nrow(a)) {
      for (j in 1:nrow(a)) {
         i1 <- rowSums(a[i,] < a[j,]) == ncol(a)
         a[i1,, drop = FALSE]
      }
   } 
   return(a) 
}
l = Dominance(df)

> l
  X1 X2
1 45 65
2 50 70

I expect the pair (45,65) to be removed.

In case it is `45 65`the pair is removed. If it is `45 85` the pair stays. — Simon, Jul 29 '19 at 14:21

akrun · Accepted Answer · 2019-07-29T14:20:35.843

1

An option is to use do a comparison (<) between equal sized objects, then get the rowSums, if the sum is equal to the number of columns of the dataset, it implies all the elements in that row is less than the second data corresponding row

f1 <- function(mat1, mat2) {
            i1 <- !rowSums(mat1 < mat2) == ncol(mat1)
            i2 <- !rowSums(mat2 < mat1) == ncol(mat2)
           rbind(mat1, mat2)[c(i1, i2),, drop = FALSE]
   }

b <- matrix(c(45,65), ncol = 2)
b1 <- matrix(c(45,85), ncol = 2)



f1(a, b)
#     [,1] [,2]
#[1,]   50   70

f1(a, b1)
#     [,1] [,2]
#[1,]   50   70
#[2,]   45   85

edited Jul 29 '19 at 14:20

answered Jul 29 '19 at 13:35

akrun

874,273
37
540
662

Looks good. How to remove the dominated pair from the set then? i.e. when the result is TRUE? – Simon Jul 29 '19 at 13:38
I am checking right now! Not sure it works, seems it is not an index that is stored in `i1`, is that possible? – Simon Jul 29 '19 at 13:56
@Simon It is a logical vector and wrapping with `which(i1)` returns the row position. – akrun Jul 29 '19 at 14:03
I think OP needs - `rbind( a[rowSums(a < b) == ncol(a), , drop = F], b[rowSums(b < a) == ncol(b), , drop = F] )` – Shree Jul 29 '19 at 14:04
@Simon Do you need to check the reverse and do an `rbind` as Shree suggessted – akrun Jul 29 '19 at 14:07
@akrun, yes indeed I need to check the reverse. Updated my question with your suggestion. Are those 2 `for` loops ok to do what Shree suggests? – Simon Jul 29 '19 at 14:14
@akrun it works perfectly! Now do you know how to loop it for all rows in a data set? – Simon Jul 29 '19 at 14:32
@Simon It should work for all rows of the dataset assuming both are equal dimensions – akrun Jul 29 '19 at 14:32
@akrun what if I have a data set with 100 lines? – Simon Jul 29 '19 at 14:58
@Simon Doesn't matter as long as both datasets have same number of rows – akrun Jul 29 '19 at 15:01
@akrun and if I only have one data set and want to check all the lines pairwise? – Simon Jul 29 '19 at 15:03
@Simon Then you could do a `combn(seq_len(nrow(a)), 2, FUN = ` – akrun Jul 29 '19 at 15:04
It says mat2 is missing – Simon Jul 29 '19 at 15:14

I there a R function for finding a minimum between to pairs?

1 Answers1