hexagons {hexbin}R Documentation

Add Hexagon Cells to Plot

Description

Plots cells in an hexbin object. The function distinquishes among counts using 5 different styles. This function is the hexagon plotting engine called from plot.hexbin().

Usage

hexagons(dat, style = c("colorscale", "centroids", "lattice",
                        "nested.lattice", "nested.centroids"),
         minarea = 0.05, maxarea = 0.8,
         mincnt = 1, maxcnt = max(dat$cnt), trans = NULL,
         colorcut = seq(1, 0, length = 17),
         density = NULL, border = NULL, pen = NULL,
         colramp = function(n){ LinGray(n,beg = 90, end = 15) },
         verbose = getOption("verbose"))

Arguments

dat an object of class hexbin, see hexbin.
style character specifying the type of plotting; must be one of ("grayscale","lattice","centroids","nested.lattice","nested.centroids").
minarea numeric, the fraction of cell area for the lowest count.
maxarea the fraction of the cell area for the largest count.
mincnt Cells with smaller counts are not shown.
maxcnt Cells with large counts are not shown.
trans a transformation function (or NULL) for the counts such as sqrt().
colorcut A vector of values covering [0, 1] which determine hexagon color class boundaries or hexagon size boundaries.
density polygon argument for shading. -1 causes the polygon not to be filled.
border polygon() argument. Draw the border for each hexagon.
pen colors for polygon(). Determines the color with which the polygon will be filled.
colramp function of an integer argument n returning n colors. n is determined
verbose logical indicating if some diagnostic output should happen.

Details

The five plotting styles have the following effect:

style="lattice" or "centroids"

Plots the hexagons in different sizes based on counts. The "lattice" version centers the hexagons at the cell centers whereas "centroids" moves the hexagon centers close to the center of mass for the cells. In all cases the hexagons will not plot outside the cell unless maxarea > 1. Counts are rescaled into the interval [0,1] and colorcuts determine the class boundaries for sizes and counts. The pen argument for this style should be a single color or a vector of colors of length(bin$cnt).

style="colorscale"
Counts are rescaled into the interval [0,1] and colorcuts determines the class boundaries for the color classes. For this style the function specified in coloramp is used to define the n colors for the n+1 color cuts. In for this style the pen argument is ignored.
style="nested.lattice" and "nested.centroids"
Counts are partioned into classes by power of 10. The encoding nests hexagon size within powers of 10 color contours.

If the pen argument is used it should be a matrix of colors with 2 columns and either ceiling(log10(max(bin$cnt))) or length(bin$cnt) rows. The default uses the R color palatte so that pens numbers 2-11 determine colors for completely filled cell Pen 2 is the color for 1's, Pen 3 is the color for 10's, etc. Pens numbers 12-21 determine the color of the foreground hexagons. The hexagon size shows the relative count for the power of 10. Different color schemes give different effects including 3-D illusions

Hexagon size encoding minarea and maxarea determine the area of the smallest and largest hexagons plotted. Both are expressed fractions of the bin cell size. Typical values might be .04 and 1. When both values are 1, all plotted hexagons are bin cell size, if maxarea is greater than 1 than hexagons will overlap. This is sometimes interesting with the lattice and centroid styles.

Count scaling

relcnt <- (trans(cnt)-trans(mincnt)) / (trans(maxcnt)-trans(mincnt))
area <- minarea + relcnt*maxarea

By default the transformation trans() is the identity function. The legend routine requires the transformation inverse for some options.

Count windowing mincnt and maxcnt Only routine only plots cells with cnts in [mincnts, maxcnts]

SIDE EFFECTS

Adds hexagons to the plot.

Author(s)

Dan Carr <dcarr@voxel.galaxy.gmu.edu>; ported and extended by Nicholas Lewin-Koh nikko@hailmail.net.

References

Carr, D. B. (1991) Looking at Large Data Sets Using Binned Data Plots, pp. 7–39 in Computing and Graphics in Statistics; Eds. A. Buja and P. Tukey, Springer-Verlag, New York.

See Also

hexbin, smooth.hexbin, erode.hexbin, hcell, hcell2xy, plot.hexbin, hboxplot, hdiffplot, hmatplot, hex.legend

Examples


set.seed(506)
x <- rnorm(10000)
y <- rnorm(10000)

# bin the points
bin <- hexbin(x,y)

## setup coordinate system:
plot(bin$xbnd, bin$ybnd, type="n", main = "Bivariate rnorm(10000)", asp=1)
hexagons(bin)

#  A better approach uses plot.hexbin
#  which controls the plot shape :
plot(bin, main = "Bivariate rnorm(10000)")
## or
plot(hexbin(x, y + x*(x+1)/4),
     main = "(X, X(X+1)/4 + Y)  where X,Y ~ rnorm(10000)")

# A mixture distribution
x <- c(rnorm(5000),rnorm(5000,4,1.5))
y <- c(rnorm(5000),rnorm(5000,2,3))
bin <- hexbin(x,y)

# Show color control and overplotting of hexagons
plot(bin$xbnd, bin$ybnd, type="n", main = "Bivariate mixture (10000)", asp=1)
hexagons(bin, style= "lattice", border = gray(.1), pen = gray(.6),
         minarea = .1, maxarea = 1.5)
## How to treat 'singletons' specially:
plot(bin$xbnd, bin$ybnd, type="n", main = "Bivariate mixture (10000)", asp=1)
hexagons(bin, style= "nested.centroids", mincnt = 2)# not the single ones
hexagons(bin, style= "centroids", maxcnt = 1, maxarea=0.04)# single points

# And if we had all the information...
if(require(gpclib)){
  h1 <- chull(x[1:5000], y[1:5000])
  h2 <- chull(x[5001:10000], y[5001:10000])
  h2 <- h2+5000
  h1 <- as(cbind(x[1:5000],y [1:5000])[h1, ], "gpc.poly")
  h2 <- as(cbind(x,y)[h2, ], "gpc.poly")
  plot(bin$xbnd, bin$ybnd, type="n", main = "Bivariate mixture (10000)", asp=1)
  plot(h1,poly.args = list(col ="#CCEBC5"),add = TRUE)
  plot(h2,poly.args = list(col ="#FBB4AE"),add = TRUE)
  plot(intersect(h1, h2), poly.args = list(col = 2), add = TRUE)
  hexagons(bin, style= "centroids", border = gray(.1), pen = gray(.6),
           minarea = .1, maxarea = 1.5)
}


[Package hexbin version 1.0.10 Index]