rhrKDE: Kernel Density Estimation (KDE)

Description Usage Arguments Details Value Author(s) See Also Examples

Description

A function to estimate home ranges with kernel density estimation.

Usage

1
2
3
4
5
  rhrKDE(xy, levels = 95, h = "href", xrange = NULL,
    yrange = NULL, increaseExtent = NULL, trast = NULL,
    buffer = NULL, ud = TRUE, cud = TRUE, res = NULL,
    rescale = "unitvar", lscvSearch = NULL,
    lscvWhichMin = "global", gridsize = NULL)

Arguments

xy

data.frame with two columns: x and y coordinates.

levels

numeric vector with the isopleth levels.

h

character ("href" or "hlscv") specifying the method to estimate the bandwidth or numeric value specifying the bandwidth.

xrange

numeric vector specifying min and max x for the output grid.

yrange

numeric vector specifying min and max y for the output grid.

increaseExtent

numeric value by which the x and y range are extended, see also ?extendrange.

trast

a RasterLayer used as an template for the output grid.

buffer

numeric value to buffer the bounding box of points in map units from which the extent of the output grid will be retrieved.

res

numeric value specifying the resolution for the output grid.

gridsize

a vector of length 2, specifying the number of rows and the number of columns for the ouput grid.

ud

logical value, indicating whether or not a utilization distribution should be calculated.

cud

logical value, indicating whether or not a cumulative utilization distribution should be calculated.

rescale

character value specifying if the data should be rescaled before calculating bandwidth. Possible values are: unitvar to rescale to unit variance, unitx to rescale data to variance of x and none.

lscvSearch

numeric vector of length 2, specifying lower and uper bound for candidate bandwidth (as portion of reference bandwidth) for estimating bandwidth with least squre cross validation.

lscvWhichMin

character value, specifying how candidate bandwidths are chosen with least squre cross validation. Possible values are: global or local minimum.

Details

The size and resolution of the resulting utilization distribution (UD) grid is influenced by traster, xrange, yrange, increaseExtent, buffer, res, gridsize. The size of the grid can be set either through a template raster (traster), xrange and yrange or increaseExtent. traster takes precedence over xrange and yrange, buffer and grid. If none of the previous arguments are provided, xrange and yrange are taken from the data.

The resolution of the resulting UD grid can be set through either res or gridsize. res takes precedence over gridsize. If none of the previous arguments is provided the grid is set by default to a 100 by 100 grid. The bandwidth can be provided by the user or estimated through the reference bandwidth (this method is often refered to as the ad hoc method), plug in the euqtion method or the least square cross validation method. Reference bandwidth estimation is implemented as suggested by Silverman 1986. Plugin the equation method is wrapped from KernSmooth::dpki and a simple binned version Silverman's suggestion for least square cross validation is implemented. Kernels densities are estimated with KernSmooth::bkde2d. This is a binned approximation of 2D kernel density estimates (see ?KernSmooth::bkde2d for more details.

Value

object of class RhrHREstimator

Author(s)

Johannes Signer

See Also

KernSmooth::bkde2d, KernSmooth::dpik, rhr::rhrHref, rhr::rhrHlscv, rhr::rhrHpi

Examples

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
data(datSH)
## Not run: 
# Kernel with href bandwidth estimation
k1 <- rhrKDE(datSH[, 2:3], h="href", res=100)
plot(k1)

# what is the actually estimated bandwidth?
k1$parameters$h

# Kernel with href bandwidth estimation
k2 <- rhrKDE(datSH[, 2:3], h="lscv", res=100)
plot(k2)

# what is the actually estimated bandwidth?
k2$parameters$h

## End(Not run)

rhr documentation built on May 2, 2019, 6:06 p.m.

Related to rhrKDE in rhr...