fields (version 1.2)

nkreg: Normal kernel regression estimate

Description

Estimates a univariate or multivariate regression surface using a normal kernel function and a fixed bandwidth.

Usage

nkreg(data.x, data.y, bandwidth, n.points=50, grid=NULL, grid.list=NULL)

Arguments

data.x

a matrix of data locations

data.y

a vector of values for smoothing

bandwidth

bandwidth for kernel if a vector then an estimate will be be found for each bandwidth.

n.points

Number of equally spaced points to evaluate a univariate estimate estimate or a 2-d estimate on a regular grid.

grid

matrix of locations to evaluate the kernel estimate

grid.list

A grid.list that describes the regular grid to evaluate the estimate. If it is missing the default for 2-d is to create a grid

Value

• x
  Points for evaluation
• y
  estimate of curve or surface. If more than one bandwidth is supplied then y is matrix with columns indexed by the bandwidth values
• h
  bandwidths used.
• grid.list
  grid.list that was either passed or created.

See Also

Tps, Krig, sreg, smooth.2d, image.smooth

Examples

out<- nkreg( rat.diet$t, rat.diet$con, bandwidth=3, n.points=100)
plot( rat.diet$t, rat.diet$con)
lines( out$x, out$y)

# 2-d example evaluate at 40 points
out<- nkreg( precip$x, precip$y, bandwidth=.5, n.points=64)
image.plot( as.surface(out$x, out$y))