bivRegion {refreg}

Bivariate reference region estimation

Description

This functions estimate a probabilistic/reference region for bivariate data. It is based on a kernel density estimation. It may be applied to a set of bivariate data points, or to a bivRegr object. In the former case, the function will estimate a bivariate reference region for the model standarized residuals.

Usage

bivRegion(
  Y = fit,
  H_choice = "Hcov",
  tau = 0.95,
  k = 20,
  display_plot = TRUE,
  shape = NULL,
  ...
)

Arguments

Y

A set of bivariate data points, or a bivRegr object.

H_choice

Kernel bandwidth selection method: "plug.in" for plug.in method, "LSCV" for least squate cross valiation, "SCV" for smooth cross validation, and "Hcov" for a bandwidth selection method which optimize the region coverage.

tau

A number or vector defining the desired coverage(s) of the bivariate reference region.

k

In case of using "Hcov" the number of k fold cross validations replicates to be performed.

display_plot

A logical indicating if plot must be displayed during "Hcov" bandwidht estimation procedure. The plot depicts region's coverage, evaluated with k fold cross validation, depending on kernel bandwidth value.

shape

Shape parameter modulating the final shape of the bivariate probabilistic/reference region by hand.

...

Additional parameters to be modified in KernSmooth::bkde2D() function by the user (e.g. gridsize).

Value

This function return a region or a set of regions containing a given percentage of bivariate data points.

References

Duong, T. (2019) ks: Kernel Smoothing. R package version 1.11.6. https://CRAN.R–project.org/package=ks.

Matt Wand (2020). KernSmooth: Functions for Kernel Smoothing Supporting Wand & Jones (1995). R package version 2.23–18. https://CRAN.R–project.org/package=KernSmooth

Examples

Y <- cbind(rnorm(100), rnorm(100))
Y <- as.data.frame(Y)
names(Y) <- c("y1", "y2")
region <- bivRegion(Y, tau = 0.95, shape = 2)
plot(region)