Evaluates the kernel sum
\(f(u, v) = (N h_x h_y)^{-1} \sum_k \phi((u - x_k) / h_x) \phi((v - y_k) / h_y)\)
on a regular grid, with kernel standard deviations
from the normal-reference plug-in rule
\(h = 1.06 \min(s, IQR / 1.34) N^{-1/5}\) (method "nrd", equal to
MASS::kde2d with its default bandwidth) or Scott's rule
\(h = s N^{-1/6}\) (method "scott"). Identical to the Python arm
morie.fn.ptkde.spatial_kde.
Usage
SpatialKde(data, method = "nrd", h = NULL, n = 25, lims = NULL)
Arguments
- data
Two-column matrix (or data frame) of point coordinates.
- method
"nrd" or "scott" ("default" means "nrd").
- h
Kernel standard deviations (scalar or pair); overrides the rule.
- n
Grid size per axis (scalar or pair).
- lims
Grid limits c(xmin, xmax, ymin, ymax); default the data range.
Value
List: x, y (grid), z (density matrix, rows along x), bandwidth, n.
References
Silverman, B. W. (1986). Density Estimation for Statistics and
Data Analysis. Chapman and Hall.
Scott, D. W. (1992). Multivariate Density Estimation. Wiley.
Venables, W. N. and Ripley, B. D. (2002). Modern Applied Statistics with
S, 4th edn. Springer.
Examples
pts <- cbind(c(0.1, 0.4, 0.5, 0.9, 0.3), c(0.2, 0.9, 0.4, 0.7, 0.6))
SpatialKde(pts, n = 3)$z
#> [,1] [,2] [,3]
#> [1,] 1.63941181 0.5507473 0.1301617
#> [2,] 0.84849560 1.5835738 1.2059057
#> [3,] 0.02546844 1.1059336 0.8210617
