space.maker: Creating multidimensional spaces

In TGuillerme/dispRity: Measuring Disparity

Creating multidimensional spaces

Description

Creates a multidimensional space with a given number of elements and dimensions

Usage

space.maker(
  elements,
  dimensions,
  distribution,
  arguments = NULL,
  cor.matrix = NULL,
  scree = NULL,
  elements.names = NULL,
  replicates = NULL
)

Arguments

elements	An numeric value.
dimensions	An numeric value smaller than elements.
distribution	One or more functions to determine the distribution of the elements along each dimension. The function must have a single input: elements.
arguments	Optional list of arguments to be passed to the distributions functions in the order they appear (default = NULL, see details).
cor.matrix	An optional correlation matrix of size dimensions * dimensions (default = NULL, see details).
scree	An optional proportional numeric vector for approximating the dimensions variance (default = NULL, see details).
elements.names	Optional, a character or integer string for naming the elements in the matrix.
replicates	Optional, an integer to replicate the simulations and generating multiple spaces.

Details

When passing additional arguments to different distributions, these must be given as a list to each function in the order they appear. For example if distribution = c(runif, rnorm, rgamma) and one wants the distributions to be runif(elements, min = 1, max = 10), rnorm(elements, mean = 8) and rgamma(elements, shape = 1, log = TRUE), the additional arguments should be passed as c(list(min = 1, max = 10), list(mean = 8), list(shape = 1, log = TRUE). If no arguments have to be passed to a certain function, it can be left as NULL (e.g. c(list(min = 1, max = 10), list(NULL), list(shape = 1, log = TRUE)).

The cor.matrix argument should be a correlation matrix between the dimensions. If not NULL, the multidimensional space is multiplied by the the Choleski decomposition (chol) of the correlation matrix. The scree argument is simply a value multiplier for each dimension to adjust their variance to approximate the scree one.

Author(s)

Thomas Guillerme

See Also

null.test, test.dispRity.

Examples

## A square space
plot(space.maker(5000, 2, runif), pch = 20)

## A circular space
plot(space.maker(5000, 2, rnorm), pch = 20)

## A 2-dimensional cylindrical space
plot(space.maker(5000, 2, c(rnorm, runif)), pch = 20)

## A 4-dimensional space with different distributions
space.maker(5, 4, c(runif, runif, rnorm, rgamma),
     arguments = list(list(min = 1, max = 10), list(min = 1, max = 2),
     list(mean = 8), list(shape = 1)))

## A 3-dimensional correlated space
cor_matrix <- matrix(cbind(1, 0.8 ,0.2, 0.8, 1, 0.7, 0.2, 0.7, 1), nrow = 3)
space <- space.maker(10000, 3, rnorm, cor.matrix = cor_matrix)
round(cor(space), 1) ; cor_matrix ## Both should be really similar matrices

## A 3-dimensional space with a priori approximated variance for each dimension
space <- space.maker(10000, 3, rnorm, scree = c(0.6, 0.3, 0.1))
## The resulting screeplot
barplot(apply(space, 2, var))

## Generate 3 2D normal spaces with rownames 
space.maker(10, 2, rnorm, elements.names = letters[1:10], replicates = 3)

## Not run: 
require(scatterplot3d)
## A cube space
scatterplot3d(space.maker(5000, 3, runif), pch = 20)

## A plane space
scatterplot3d(space.maker(5000, 3, c(runif, runif, runif),
     arguments = list(list(min = 0, max = 0), NULL, NULL)), pch = 20)

## A sphere space
scatterplot3d(space.maker(5000, 3, rnorm), pch = 20)

## A 3D cylindrical space
scatterplot3d(space.maker(5000, 3, c(rnorm, rnorm, runif)), pch = 20)

## Generating a doughnut space
doughnut <- space.maker(5000, 3, c(rnorm, random.circle),
     arguments = list(list(mean = 0), list(runif, inner = 0.5, outer = 1)))
## Reodering the axis for projecting the dougnut in 2D
scatterplot3d(doughnut[,c(2,1,3)], pch = 20)

## End(Not run)

TGuillerme/dispRity documentation built on Dec. 21, 2024, 4:05 a.m.