Infix operator to apply functions to matrix-like objects

The infix operator %of% is a generic function which applies functions to objects. This page describes the function and the methods defined in package mixAR.

Usage

e1 %of% e2
e1 %of% e2

Arguments

e1: usually a function, the name of a function, a character vector, or a list of functions, see Details.
e2: an object, usually matrix-like.

Details

%of% is a generic function with dispatch on both arguments. It is intended to be used mainly in infix form.

%of% transforms each “column” of a matrix-like object by a function. If e1 specifies a single function, that is applied to all columns. Otherwise length(e1) should equal the number of “columns” of e2 and e1[[i]] is applied to the i-th “column” of e2.

The mental model is that the first argument, e1, is (converted to) a list of functions containing one function for each column of e2. The i-th function is applied to each element of the i-th column.

The methods for "MixComp" objects allow for very transparent and convenient computing with "MixAR" objects.

Value

for the default method, a matrix;

for methods with e2 from class MixComp, a MixComp object with its slot m replaced by the result of applying e1 to its elements, see the descriptions of the individual methods for details;

Methods

Below are the descriptions of the methods for %of% defined by package mixAR.

signature(e1 = "ANY", e2 = "ANY")

This is the default method. It uses apply() to evaluate e1 for each element of the matrix e2, without checking the arguments. If the arguments are not suitable for apply(), any error messages will come from it. So, for this method e1 is a function (or the name of a function) and e2 is a matrix or array.

signature(e1 = "function", e2 = "MixComp")

Create (and return) a MixComp object with its slot m replaced by the result of applying the function e1 to each element of the MixComp object e2, see class "MixComp".

signature(e1 = "character", e2 = "MixComp")

Here e1 contains the names of one or more functions. If length(e1) = 1, this is equivalent to the method for e1 of class "function".

If length(e1) > 1, then for each i the function specified by e1[i] is applied to the ith column of e2@m. In this case there is no recycling: e1 must have ncol(e2@m) elements.

signature(e1 = "list", e2 = "MixComp")

Here each element of e1 is a function or the name of a function. It works analogously to the method with e1 from class "character". If length(e1) = 1, then e1[[1]] is applied to each element of e1@m. Otherwise, if length(e1) > 1, then e1[[i]] is applied to the ith column of e2@m.

Author

Georgi N. Boshnakov

Note

The code is rather inefficient for some of the methods.

Maybe should require that the functions in the first argument are vectorised. (Some methods effectively assume it.)

Examples

m <- matrix(rnorm(18), ncol = 3)
## defult method
pm1 <- pnorm %of% m
f3 <- list(pnorm, function(x, ...) pnorm(x, mean = 0.1),
                  function(x, ...) pnorm(x, mean = -0.1) )
## no method for f from "list" yet:
## pm2 <- f3 %of% m

mc <- new("MixComp", m = m)
pnorm %of% mc
#> An object of class "MixComp"
#> Slot "m":
#>           [,1]      [,2]      [,3]
#> [1,] 0.6331502 0.1160244 0.6740943
#> [2,] 0.4306943 0.1910889 0.3883076
#> [3,] 0.5201271 0.5757670 0.9257244
#> [4,] 0.3059220 0.2870982 0.2739692
#> [5,] 0.2498748 0.3089111 0.9584489
#> [6,] 0.8040861 0.1328869 0.9665520
#> 
pmc3 <- f3 %of% mc
## result is equivalent to applying f3[[i] to m[ , i]:
all.equal(pmc3@m, cbind(f3[[1]](m[ , 1]), f3[[2]](m[ , 2]), f3[[3]](m[ , 3])))
#> [1] TRUE