Pairwise Calculations

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

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 to be calculated. It is assumed, that the first 2 arguments denominate x and y.

...

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 transposed. Default is FALSE.

Details

This code is based on the logic of cor() and extended for asymmetric functions.

Value

a matrix with the results of FUN.

Author

Andri Signorell <andri@signorell.net>

Examples

PairApply(d.diamonds[,c("colour","clarity","cut","polish")], FUN = CramerV, 
          symmetric=TRUE)
#>            colour   clarity       cut    polish
#> colour  1.0000000 0.2385027 0.1716311 0.1749971
#> clarity 0.2385027 1.0000000 0.2726831 0.2842774
#> cut     0.1716311 0.2726831 1.0000000 0.2196234
#> polish  0.1749971 0.2842774 0.2196234 1.0000000

# user defined functions are ok as well
PairApply(d.diamonds[,c("clarity","cut","polish","symmetry")], 
  FUN = function(x,y) wilcox.test(as.numeric(x), as.numeric(y))$p.value, symmetric=TRUE)
#>               clarity          cut       polish     symmetry
#> clarity  1.000000e+00 1.528415e-12 1.381603e-37 1.420752e-40
#> cut      1.528415e-12 1.000000e+00 1.702690e-18 2.090998e-21
#> polish   1.381603e-37 1.702690e-18 1.000000e+00 2.307068e-01
#> symmetry 1.420752e-40 2.090998e-21 2.307068e-01 1.000000e+00

# asymetric measure
PairApply(d.diamonds[,c("colour", "clarity", "cut", "polish")], 
  FUN = Lambda, direction = "row")
#>             colour    clarity        cut     polish
#> colour  1.00000000 0.10526316 0.03047091 0.03878116
#> clarity 0.11111111 1.00000000 0.10802469 0.08333333
#> cut     0.02405498 0.05154639 1.00000000 0.04810997
#> polish  0.06779661 0.18644068 0.10593220 1.00000000

# ... compare to:
Lambda(x=d.diamonds$colour, y=d.diamonds$clarity, direction="row")  
#> [1] 0.1052632
Lambda(x=d.diamonds$colour, y=d.diamonds$clarity, direction="column")  
#> [1] 0.1111111


# the data.frame
dfrm <- d.diamonds[, c("colour","clarity","cut","polish")]
PairApply(dfrm, FUN = CramerV, symmetric=TRUE)
#>            colour   clarity       cut    polish
#> colour  1.0000000 0.2385027 0.1716311 0.1749971
#> clarity 0.2385027 1.0000000 0.2726831 0.2842774
#> cut     0.1716311 0.2726831 1.0000000 0.2196234
#> polish  0.1749971 0.2842774 0.2196234 1.0000000

# the same as matrix (columnwise)
m <- as.matrix(dfrm)
PairApply(m, FUN = CramerV, symmetric=TRUE)
#>            colour   clarity       cut    polish
#> colour  1.0000000 0.2385027 0.1716311 0.1749971
#> clarity 0.2385027 1.0000000 0.2726831 0.2842774
#> cut     0.1716311 0.2726831 1.0000000 0.2196234
#> polish  0.1749971 0.2842774 0.2196234 1.0000000

# ... and the list interface
lst <- as.list(dfrm)
PairApply(lst, FUN = CramerV, symmetric=TRUE)
#>            colour   clarity       cut    polish
#> colour  1.0000000 0.2385027 0.1716311 0.1749971
#> clarity 0.2385027 1.0000000 0.2726831 0.2842774
#> cut     0.1716311 0.2726831 1.0000000 0.2196234
#> polish  0.1749971 0.2842774 0.2196234 1.0000000