Package 'hexDensity'

Title: Fast Kernel Density Estimation with Hexagonal Grid
Description: Kernel density estimation with hexagonal grid for bivariate data. Hexagonal grid has many beneficial properties like equidistant neighbours and less edge bias, making it better for spatial analyses than the more commonly used rectangular grid. Carr, D. B. et al. (1987) <doi:10.2307/2289444>. Diggle, P. J. (2010) <doi:10.1201/9781420072884>. Hill, B. (2017) <https://blog.bruce-hill.com/meandering-triangles>. Jones, M. C. (1993) <doi:10.1007/BF00147776>.
Authors: Quoc Hoang Nguyen [aut, cre] , Yunshun Chen [aut] , Dan Carr [ctb], binning routine ported by Nicholas Lewin-Koh [ctb], Martin Maechler [ctb]
Maintainer: Quoc Hoang Nguyen <[email protected]>
License: GPL
Version: 1.4.8
Built: 2025-03-13 05:01:44 UTC
Source: https://github.com/chenlaboratory/hexdensity

Help Index

Draw hexagon tiles with grid package

Description

Adapted from grid.hexagons by hexbin with speedup specific for hexagonal tiling (avoid plotting the most abundance hexagons by setting its color as background).

Usage

grid.hexagontile(
  hexDensity,
  use.count = TRUE,
  cell.at = NULL,
  trans = NULL,
  colorcut = seq(0, 1, length = 1024),
  colramp = colorRampPalette(col.viridis),
  def.unit = "native"
)

Arguments

hexDensity

hexbin object returned by hexDensity.
use.count

logical specifying if counts from hexbin object should be used.
cell.at

numeric vector to be plotted instead of counts, must be same length as the number of cells.
trans

a transformation function (or NULL) for the counts, e.g., sqrt.
colorcut

An integer for the number of equi-spaced colorcut in [0,1] to assign colors to values. Alternatively, a vector of custom colorcut spacing between [0, 1].
colramp

Color function that accept an integer n and return n colors.
def.unit

Default unit to be used.

Value

No return value

SIDE EFFECTS

Adds hexagons to plot

Author(s)

Dan Carr <[email protected]>; ported and extended by Nicholas Lewin-Koh [email protected]. Modified by Quoc Hoang Nguyen <[email protected]> for hexDensity.

Hexagonal binning with whole grid output.

Description

Adapted from hexbin to output hexagons with 0 count, and also to allow points with different weights. Default to use regular hexagon. See hexbin for extra detail on the output.

Usage

hexbinFull(
  x,
  y = NULL,
  xbins = 128,
  shape = NULL,
  xbnds = range(x),
  ybnds = range(y),
  xlab = NULL,
  ylab = NULL,
  IDs = FALSE,
  weight = NULL
)

Arguments

x, y

Coords of the points or a single plotting structure to be used in binning. See xy.coords.
xbins

Number of bins in a row.
shape

shape = yheight/xwidth of the plotting regions
xbnds, ybnds

Horizontal and vertical limits of the binning region in x or y units respectively, must encompass range(x) or range(y) respectively; Vector of length 2
xlab, ylab

Optional character strings used as labels for x and y. If NULL, sensible defaults are used.
IDs

Logical indicating if the hexagonal cell ID for each point should be returned, see hexbin.
weight

Numeric weight vector to be assigned to points.

Value

an S4 object of class hexbin.

Author(s)

Dan Carr <[email protected]>; ported and extended by Nicholas Lewin-Koh <[email protected]>. Modified by Quoc Hoang Nguyen <[email protected]> for hexDensity.

References

Carr, D. B. et al. (1987) Scatterplot Matrix Techniques for Large N. JASA 83, 398, 424–436.

Examples

set.seed(133)
d=hexbinFull(x=rnorm(20000),y=rnorm(20000),xbins=50)
plotHexDensity(d)

Generate contour for a hexagonal grid.

Description

Algorithm is a modification of the meandering triangles as described in https://blog.bruce-hill.com/meandering-triangles to work with hexagons. See isolines for details about the output.

Usage

hexContour(hexDensity, levels)

Arguments

hexDensity

hexDensity object to be contoured.
levels

Numeric vector for which contour lines should be generated

Details

This function is made to follow the same behaviour as isolines. A dedicated plotting function is in the work. Meanwhile, see example of how to plot the output with ggplot2's geom_path.

Value

A list of x, y, and ID, for the contour line at each levels. ID indicates the different line segments making up the contour.

Examples

set.seed(133)
x=rnorm(200)
y=rnorm(200)
d = hexDensity(x=x,y=y,bandwidth=0.4)
cutoff=quantile(d@count,0.9)
lines = hexContour(d,cutoff)

library(ggplot2)
library(hexbin)
#plot against density
ggplot()+
  geom_point(
    aes(x=hcell2xy(d)$x,
        y=hcell2xy(d)$y,
        col=d@count)
 ) +
  scale_color_viridis_c()+
  geom_path(
    aes(
     x = lines[[1]]$x, y = lines[[1]]$y, group = lines[[1]]$id
    )
  )

#plot against data points
ggplot() +
  geom_point(
    aes(x=x,y=y)) +
  geom_path(
    aes(
      x = lines[[1]]$x, y = lines[[1]]$y, group = lines[[1]]$id
    )
  )

Kernel Density Estimation with Hexagonal grid.

Description

Kernel Density Estimation with Hexagonal grid.

Usage

hexDensity(
  x,
  y = NULL,
  xbins = 128,
  bandwidth = NULL,
  edge = TRUE,
  diggle = FALSE,
  weight = NULL,
  ...
)

Arguments

x, y

Coords of the points or a single plotting structure to be used in binning. See xy.coords.
xbins

Number of bins in a row.
bandwidth

Bandwidth for the smoothing kernel. Either a scalar, a vector of length 2, or a 2x2 variance-covariance matrix for the bandwidths in the x and y directions.
edge

Logical value for whether to apply edge correction. Default is TRUE.
diggle

Logical value for apply edge correction with the more accurate Jones-Diggle method (need 'edge' to be TRUE).
weight

numeric weight vector to be assigned to points.
...

arguments for hexbinFull

Details

Default bandwidth is 1/8 of the range of the smaller dimensions.

Value

an S4 object of class hexbin.

References

Diggle, P. J. (2010) Nonparametric methods. Chapter 18, pp. 299–316 in A.E. Gelfand, P.J. Diggle, M. Fuentes and P. Guttorp (eds.) Handbook of Spatial Statistics, CRC Press, Boca Raton, FL.

Jones, M. C. (1993) Simple boundary corrections for kernel density estimation. Statistics and Computing 3, 135–146.

Examples

set.seed(133)
d = hexDensity(x=rnorm(200),y=rnorm(200),bandwidth=0.15)

Meandering triangles for hexagonal grid in C++

Description

Meandering triangles for hexagonal grid in C++

Usage

meanderingTriangles(x.coords.left, x.coords.right, y.coords, z, levels)

Arguments

x.coords.left

Vector for x coords of left-aligned rows (row 1,3,5,...)
x.coords.right

Vector for x coords of right-aligned rows (row 2,4,6,...)
y.coords

Vector for y coords of all rows.
z

Matrix for elevation values for the grid point
levels

Vector of z value cutoffs for contouring.

Details

This function is not meant to be used as is, unless you are very familiar with how hexContour works.

Value

list of x, y, and ID, for the contour line at each levels.

References

Hill, B. (2017) Meandering triangles. Naming Things. https://blog.bruce-hill.com/meandering-triangles

Plotting method for hexagonal Kernel Density Estimation

Description

Adapted the plotting function from hexbin. X and Y axes now have the same scale with option for different aspect ratio. Ribbon legend for continuous data.

Usage

plotHexDensity(
  hexDensity,
  main = NULL,
  xlab = NULL,
  ylab = NULL,
  xaxt = TRUE,
  yaxt = TRUE,
  lcex = 1,
  colramp = colorRampPalette(col.viridis),
  colorcut = 1024,
  legend = TRUE,
  legendWidth = 0.05,
  legendDistance = 0.15,
  aspectRatio = diff(hexDensity@xbnds)/diff(hexDensity@ybnds),
  margin = 0.18,
  newpage = TRUE
)

Arguments

hexDensity

hexbin object returned by hexDensity
main

Main title
xlab, ylab

x-axis and y-axis label
xaxt, yaxt

Logical. Whether to plot x,y axes
lcex

Expansion factor for all letters.
colramp

Color function that accept an integer n and return n colors.
colorcut

An integer for the number of equi-spaced colorcut in [0,1] to assign colors to values. Alternatively, a vector of custom colorcut spacing between [0, 1].
legend

Legend is currently non-functional and should be ignored.
legendWidth

Expansion factor for legend width.
legendDistance

Expansion factor for the space between the plot and the legend.s
aspectRatio

width to height ratio of the plot. Default is the (inverse of) shape value of hexDensity.
margin

Minimum guaranteed margin for the plot. Different aspect ratio between the screen and the plot means that margin can be larger on certain sides.
newpage

logical for whether to plot on a new page.

Value

No return value

SIDE EFFECTS

Create kernel density estimate plot with hexagons

Author(s)

Dan Carr <[email protected]>; ported and extended by Nicholas Lewin-Koh [email protected]. Modified by Quoc Hoang Nguyen <[email protected]> for hexDensity.

Examples

set.seed(133)
d = hexDensity(x=rnorm(200),y=rnorm(200),bandwidth=0.15)
plotHexDensity(d)

Find the hexagon cells from xy coordinates given a hexbin object.

Description

Find the hexagon cells IDs from xy coordinates given a hexbin object. Useful if you want to get the KDE value at a certain coordinate.

Usage

xy2hcell(
  hexbin = NULL,
  x,
  y = NULL,
  xbins = NULL,
  xbnds = NULL,
  ybnds = NULL,
  shape = NULL
)

Arguments

hexbin

hexbin object to be referenced to.
x, y

coordinates or vectors of coordinates of the points.
xbins

number of bins partitioning the range of xbnds
xbnds, ybnds

horizontal and vertical limit of the binning region. Must be numeric vector of length 2.
shape

shape = yheight/xwidth of the plotting regions.

Details

If a hexbin object is not provided, parameters of the binning region (xbins, xbnds, ybnds, shape) can be used instead. For finding the xy coordinates of the hexagons for a hexbin object, see hcell2xy.

Value

a vector the same length as x with the hexagonal cell ID for each point

Examples

library(hexbin)
set.seed(133)
d=hexDensity(x=rnorm(20000),y=rnorm(20000),xbins=50)
#Get KDE value at the coordinate x=0,y=0
loc = xy2hcell(d,x=0,y=0)
d@count[loc]