as.covar: as.covar

In TGuillerme/dispRity: Measuring Disparity

as.covar

Description

Changes a dispRity metric to use the covar element from a dispRity object.

Usage

as.covar(fun, ..., VCV = TRUE, loc = FALSE)

Arguments

fun	a function to apply to the $covar element of dispRity.
...	any additional arguments to pass to fun.
VCV	logical, whether to use the $VCV component of the elements in dispRity$covar (TRUE; default) or not (FALSE) (see details).
loc	logical, whether to use the $loc component of the elements in dispRity$covar (TRUE) or not (FALSE; default) (see details).

Details

This function effectively transforms the input argument from matrix (or matrix2) to matrix = matrix$VCV and adds a evaluation after the return call to indicate that the function works on a $covar element. Note that if the function does not have an argument called matrix, the first argument is estimated as being the one to be transformed (e.g. if the function has its first argument x, it will transform it to x = x$VCV).

You can toggle between using the $VCV or the $loc argument in the $covar matrix by using either VCV = TRUE, loc = FALSE (to access only fun(matrix = matrix$VCV, ...)), VCV = FALSE, loc = TRUE (to access only matrix = matrix(matrix$loc, nrow = 1), ...) or VCV = TRUE, loc = TRUE (to access fun(matrix = matrix$VCV, loc = matrix$loc, ...); provided fun has an extra loc argument).

For between.groups metrics with matrix and matrix2 arguments, you can provide multiple logicals for VCV and loc to be applied repspectively to matrix and matrix2. For example VCV = TRUE will reinterpret matrix and matrix2 as matrix$VCV and matrix2$VCV but loc = c(TRUE, FALSE) will only reinterpret matrix as matrix$loc (and matrix2 will not be reinterpreted).

Author(s)

Thomas Guillerme

See Also

dispRity MCMCglmm.subsets

Examples

## Creating a dispRity
data(charadriiformes)

## Creating a dispRity object from the charadriiformes model
covar_data <- MCMCglmm.subsets(data       = charadriiformes$data,
                               posteriors = charadriiformes$posteriors)

## Get one matrix and one covar matrix
one_matrix <- get.matrix(covar_data, subsets = 1)
one_covar  <- get.covar(covar_data, subsets = 1, n = 1)[[1]][[1]]

## Measure the centroids
centroids(one_matrix)

## Measure the centroids on the covar matrix
as.covar(centroids)(one_covar)
## Is the same as:
centroids(one_covar$VCV)

## Apply the measurement on a dispRity object:
## On the traitspace:
summary(dispRity(covar_data, metric = c(sum, centroids))) 
## On the covariance matrices:
summary(dispRity(covar_data, metric = c(sum, as.covar(centroids))))
## The same but with additional options (centre = 100)
summary(dispRity(covar_data,
                 metric = c(sum, as.covar(centroids)),
                 centre = 100))

## Example with the VCV and loc options
## A metric that works with both VCV and loc
## (the sum of the variances minus the distance from the location)
sum.var.dist <- function(matrix, loc = rep(0, ncol(matrix))) {
    ## Get the sum of the diagonal of the matrix
    sum_diag <- sum(diag(matrix))
    ## Get the distance between 0 and the loc
    dist_loc <- dist(matrix(c(rep(0, ncol(matrix)), loc), nrow = 2, byrow = TRUE))[1]
    ## Return the sum of the diagonal minus the distance
    return(sum_diag - dist_loc)
}
## Changing the $loc on one_covar for the demonstration
one_covar$loc <- c(1, 2, 3)
## Metric on the VCV part only
as.covar(sum.var.dist, VCV = TRUE, loc = FALSE)(one_covar)
## Metric on the loc part only
as.covar(sum.var.dist, VCV = FALSE, loc = TRUE)(one_covar)
## Metric on both parts
as.covar(sum.var.dist, VCV = TRUE, loc = TRUE)(one_covar)

TGuillerme/dispRity documentation built on April 17, 2024, 10 p.m.