Computes and draws density contour lines.

## Usage

```
ternary_density(x, y, z, ...)
# S4 method for numeric,numeric,numeric
ternary_density(
x,
y,
z,
h = NULL,
n = 25,
nlevels = 10,
levels = NULL,
palette = function(i) grDevices::hcl.colors(i, "YlOrRd", rev = TRUE),
...
)
# S4 method for ANY,missing,missing
ternary_density(
x,
h = NULL,
n = 25,
nlevels = 10,
levels = NULL,
palette = function(i) grDevices::hcl.colors(i, "YlOrRd", rev = TRUE),
...
)
```

## Arguments

- x, y, z
A

`numeric`

vector giving the x, y and z ternary coordinates of a set of points. If`y`

and`z`

are missing, an attempt is made to interpret`x`

in a suitable way (see`grDevices::xyz.coords()`

).- ...
Further arguments to be passed to

`ternary_lines()`

.- h
A length-one

`numeric`

vector giving the bandwidth.- n
A length-one

`numeric`

specifying the number of grid points.- nlevels
A length-one

`numeric`

vector specifying the number of contour levels desired. Only used if`levels`

is`NULL`

.- levels
A

`numeric`

vector of levels at which to draw contour lines.- palette
A color palette

`function`

that takes a single integer argument (the number of levels) and returns a vector of colors.

## Value

`ternary_density()`

is called it for its side-effects.

Invisibly returns a `list`

with elements `levels`

(the contour levels) and
`colors`

(the contour colors) that can be used for a legend.

## Details

Two-dimensional kernel density estimation with an axis-aligned bivariate normal kernel. Normal kernel is evaluated on a square grid, after an isometric log ratio transformation of the original data.

## Note

Two-dimensional kernel density estimation is adapted from `MASS::kde2d()`

.

**This must be considered as experimental and subject to major changes
in a future release.**

## See also

Other statistics:
`ternary_contour()`

,
`ternary_ellipse()`

,
`ternary_hull()`

,
`ternary_mean()`

,
`ternary_pca()`

## Examples

```
## Add density
## Data from Aitchison 1986
ternary_plot(lava, panel.first = ternary_grid())
levels <- ternary_density(lava, n = 500, nlevels = 10)
## Add a legend
legend_image <- grDevices::as.raster(rev(levels$colors))
graphics::rasterImage(legend_image, 0.85, 0.75, 0.9, 1)
graphics::text(x = 0.9, y = c(0.75, 1), labels = range(levels$levels), pos = 4)
```