ridgetorus (version 1.0.3)

torus_pairs: Toroidal pairs plot

Description

Pairs plots for data on \([-\pi, \pi)^d\), \(d\geq 2\). The diagonal panels contain kernel density estimates tailored to circular data.

Usage

torus_pairs(
  x,
  max_dim = 10,
  columns = NULL,
  col_data = 1,
  ylim_dens = c(0, 1.5),
  bwd = "ROT",
  scales = rep(pi, ncol(x))
)

Value

A ggplot. The density plots show the Fréchet means (red bars) and the Fréchet standard deviations (gray text).

Arguments

x

a matrix of size c(nx, d) with angles on \([-\pi, \pi)\).

max_dim

the maximum number of scores to produce the scores plot. Defaults to 10.

columns

if specified, the variables to be plotted. If NULL (the default), the first max_dim variables are plotted.

col_data

color(s) for the data points. Defaults to 1.

ylim_dens

common ylim for the diagonal plots. Defaults to c(0, 1).

bwd

type of bandwidth selector used in the kernel density plots. Either "ROT", "EMI", "AMI", "LSCV", or "LCV". See bw_dir_pi and bw_dir_cv. Defaults to "ROT".

scales

scales of the torus. Defaults to rep(pi, ncol(x)).

Details

The default bandwidth selector is the Rule-Of-Thumb (ROT) selector in García-Portugués (2013). It is fast, yet it may oversmooth non-unimodal densities. The EMI selector gives more flexible fits.

References

García-Portugués, E. (2013). Exact risk improvement of bandwidth selectors for kernel density estimation with directional data. Electronic Journal of Statistics, 7:1655--1685. tools:::Rd_expr_doi("10.1214/13-ejs821")

Examples

# Generate data
n <- 50
set.seed(123456)
x <- sdetorus::toPiInt(rbind(
  mvtnorm::rmvnorm(n = n, mean = c(-pi, -pi) / 2,
                   sigma = diag(0.1, nrow = 2)),
  mvtnorm::rmvnorm(n = n, mean = c(-3 * pi / 2, 0) / 2,
          sigma = diag(0.1, nrow = 2)),
  mvtnorm::rmvnorm(n = n, mean = c(0, pi / 2),
                   sigma = diag(0.1, nrow = 2))
))
col <- rainbow(3)[rep(1:3, each = n)]

# Torus pairs
torus_pairs(x, col_data = col)
# \donttest{
fit <- ridge_pca(x = x)
torus_pairs(fit$scores, col_data = col)# }