sspline (version 0.1-3)

smooth.sspline: Smoothing Spline on the Sphere

Description

It fits a smoothing splines on the sphere with the smoothing parameter chosen by the generalized cross validation (GCV) criteria or given by the user.

Usage

smooth.sspline(lon, lat, y, m = 2, smth = 0, lambda = 0)

Arguments

lon

numeric vector, the longitudes

lat

numeric vector, the latitudes

y

numeric vector, the observations at (lon, lat)

m

integer, order of smoothing, takes value from 1 to 10. Default to 2

smth

method for choosing the smoothing parameter: 0, gcv method; 1, user specified. Default to 0

lambda

used only when smth = 1.

Value

• A smooth.sspline object with the components
• lon
  the original longitude
• lat
  the original latitude
• obs
  the original observation
• lambda
  the lambda that minimizes the gcv score
• gcv
  the corresponding gcv value at lambda
• varhat
  the estimated variance
• c
  the coefficient vector c for the estimated function
• d
  the coefficient d for the estimated function
• yhat
  the estimated (smoothed) observation
• call
  the call to smooth.sspline

Details

It calls Fortran subroutine with the .Fortran interface.

References

Grace Wahba (1981), Spline Interpolation and Smoothing on the Sphere, SIAM J. SCI. STAT. COMPUT.

Examples

data(WTdiff)

subdat <- WTdiff[sample(nrow(WTdiff), 200), 2:4]
attach(subdat)

smooth.sspline(lon, lat, avgd)

detach(subdat)