Pairwise Calculations

Implements a logic to run pairwise calculations on the columns of a data.frame or a matrix.

Usage

pairApply(x, FUN = NULL, ..., symmetric = FALSE)

Arguments

x

a list, a data.frame or a matrix with columns to be processed pairwise.

FUN

a function (or the name of a function) to be calculated. It is assumed, that the first 2 arguments denominate x and y, and that it returns a single numeric value.

...

the dots are passed to FUN.

symmetric

logical. Does the function yield the same result for FUN(x, y) and FUN(y, x)?
If TRUE just the lower triangular matrix is calculated and mirrored. Default is FALSE.

Value

a matrix with the results of FUN.

Details

This code is based on the logic of cor() and extended for asymmetric functions. Cell [i, j] of the result contains FUN(x[[i]], x[[j]], ...), so the first argument of FUN corresponds to the row variable and the second to the column variable.

See also

outer(), pairwise.table()

Other combinatorics: combN(), combPairs(), combSet(), permn(), randGroupSplit(), sampleX()

Examples

# build a dataset
set.seed(1)
d.sub <- transform(
  data.frame(
    X1 = rnorm(n <- 300),
    X3 = rnorm(n)),
  X2 = 0.8*X1 + rnorm(n),
  X4 = 0.5*X3 + rnorm(n)
  )

pairApply(d.sub, FUN = cor, method="spearman")
#>            X1         X3          X2          X4
#> X1 1.00000000 0.03799909  0.55466483  0.01294681
#> X3 0.03799909 1.00000000  0.06833009  0.44778009
#> X2 0.55466483 0.06833009  1.00000000 -0.02556606
#> X4 0.01294681 0.44778009 -0.02556606  1.00000000

# user defined functions are ok as well
pairApply(d.sub,
  FUN = function(x,y)
    wilcox.test(as.numeric(x), as.numeric(y))$p.value, symmetric=TRUE)
#>           X1        X3        X2        X4
#> X1 1.0000000 0.6297484 0.4566135 0.4158320
#> X3 0.6297484 1.0000000 0.7499997 0.7142069
#> X2 0.4566135 0.7499997 1.0000000 0.9359926
#> X4 0.4158320 0.7142069 0.9359926 1.0000000