drvkde {feature} | R Documentation |
Compute kernel density derivative estimates and standard errors for multivariate data.
drvkde(x, drv, bandwidth, gridsize, range.x, binned=FALSE, se=TRUE)
x |
data matrix or matrix of binning counts |
drv |
vector of derivative indices |
bandwidth |
vector of bandwidths |
gridsize |
vector of grid sizes |
range.x |
list of vector of ranges for x |
binned |
flag to indicate: TRUE = x is binned counts or FALSE = x is data matrix. Default is TRUE |
se |
flag for computing standard error of kernel estimate. Default is TRUE |
The estimates and standard errors are computed over a grid of binned counts
x.grid
. If the binned counts are not supplied then they are computed
inside this function.
If gridsize
and range.x
are not supplied, they are
computed inside this function.
This function doesn't need to be used directly as it called from
featureSignif
.
Returns a list with fields
x.grid
- grid points
est
- kernel estimate of partial derivative of density function
indicated by drv
se
- estimate of standard error of est
(if se=TRUE
).
M.P. Wand wand@maths.unsw.edu.au
Wand, M.P. and Jones, M.C. (1995) Kernel Smoothing Chapman and Hall.
## univariate data(earthquake) eq3 <- -log10(-earthquake[,3]) fhat <- drvkde(x=eq3, drv=0, bandwidth=0.1) ## KDE of f fhat1 <- drvkde(x=eq3, drv=1, bandwidth=0.1) ## KDE of df/dx ## trivariate data(earthquake) earthquake[,3] <- -log10(-earthquake[,3]) fhat <- drvkde(x=earthquake, drv=c(0,0,0), bandwidth=c(0.04,0.04,0.05), se=FALSE) ## KDE of f fhat212 <- drvkde(x=earthquake, drv=c(2,1,2), bandwidth=c(0.04,0.04,0.05), se=FALSE) ## KDE of d^3 f/ dx^2 dy dz^2