aec: Analysis of Ellipsoid Coverage

In mlcollyer/ACHC: Analysis of Convex Body Coverage

Description

Function obtains relative frequencies of ellipsoid coverage in the represented data space, and uses a permutation procedure to generate confidence limits for random assignmnet of observations to groups.

Usage

aec(
  dat,
  std = FALSE,
  group,
  grid.points = 500,
  grid.space = 0.05,
  confidence = NULL,
  ellipse.density = 120,
  iter = 99,
  seed = NULL,
  print.progress = TRUE
)

Arguments

dat	Either a data frame or matrix of data to be analyzed. Because of computational limitations, the number of variables is currently limited to 5.
std	A logical value that if TRUE finds standard deviates of the data (data are both centered and scaled by variable standard deviations).
group	A factor or vector coerible to factor for defining groups.
grid.points	The desired number of points, sampled from a uniform distribution of points in the data space, within a convex hull for all observed points. This number might be less than the maximum possible number of points (which could be huge).
grid.space	The approximate spacing of uniform points along each axis. For example, 0.05 means points will be placed at increments that are 5 percent of the expanse of data, per axis.
confidence	The confidence level for ellipsoids, based on the covariance matrix of data by groups. Multivariate normality is assumed in estimation. If NULL, then ellipsoids merely reflect the span of eigenvalues for the covariance matrix of the data. Otherwise, the value should be between 0.01 and 1.
ellipse.density	A numeric value to indicate how many discrete points (in a circle) are used to approximate the continuous ellipse function. More points mean a more precise curve, but increase computation time. The default, 120 points, is the same as 3 degrees (pi/60 radians) increments.
iter	The number of iterations (permutations) to run for the test. Because the observed case counts as one iteration, this should be the number desired, minus one.
seed	Change the random seed, if desired. If NULL, the seed will equal the number of permutations.
print.progress	A logical value to indicate if permutation progress should be printed to the screen. This is useful for analyses that will run a long time.

Details

A description is needed here

Value

An object of class achc is a list containing the following

grid	The grid points obtained.
analysis	A matrix of 0s and 1s for convex hull presenece at grid points (rows) by groups (columns) for each permutation.
group	The factor of group levels used; useful for downstream functions.
perms	The number of permutations.
perm.schedule	The sampling frames in each permutation.
std	Whether data were standardized.

Author(s)

Michael Collyer

Examples

# 3 dimensions of data
library(RRPP)
data("Pupfish")
group <- interaction(Pupfish$Sex, Pupfish$Pop)
P <- prcomp(Pupfish$coords)$x[,1:3] # first 3 PCs

pupCHC <- aec(P, std = FALSE, group = group, iter = 99, confidence = 0.95,
 ellipse.density = 60, grid.points = 100, grid.space = 0.1)
pupCHC
summary(pupCHC, confidence = 0.95)
plot(pupCHC, lwd = 2)
plot(pupCHC, lwd = 2, confidence = 0.99)

mlcollyer/ACHC documentation built on May 30, 2020, 10:26 p.m.