stdize {ffmanova}

Centering and scaling of matrices

Description

Function to center and/or scale the coloumns of a matrix in various ways. The coloumns can be centered with their means or with supplied values, and they can be scaled with their standard deviations or with supplied values.

Usage

stdize(x, center = TRUE, scale = TRUE, avoid.zero.divisor = FALSE)

stdize3(x, center = TRUE, scale = TRUE, avoid.zero.divisor = FALSE)

Arguments

x

A matrix.

center

A logical, or a numeric vector. The values to subtract from each column. If center is TRUE, the mean values are used.

scale

A lgical, or a numeric vector. The values to divide each column with. If scale is TRUE, the standard deviations are used.

avoid.zero.divisor

A logical. If TRUE, each occurence of 0 in scale is replaced with a 1.

Details

stdize standardizes the coloumns of a matrix by subtracting their means (or the supplied values) and dividing by their standard deviations (or the supplied values).

If avoid.zero.divisor is TRUE, division-by-zero is guarded against by substituting any 0 in center (either calculated or supplied) with 1 prior to division.

The main difference between stdize and scale is that stdize divides by the standard deviations even when center is not TRUE.

Value

A matrix.

Note

stdize3 is a variant with a three-element list as output (x, center, scale) and where avoid.zero.divisor is also used to avoid centring (constant term in model matrix is unchanged).

Author(s)

Bjørn-Helge Mevik and Øyvind Langsrud

See Also

scale

Examples

A <- matrix(rnorm(15, mean = 1), ncol = 3)
stopifnot(all.equal(stdize(A), scale(A), check.attributes = FALSE))

## These are different:
stdize(A, center = FALSE)
scale(A, center = FALSE)