R/idw.R

In jskoien/intamap: Procedures for Automated Interpolation

preProcess.idw = function(object, ...) {
	# perhaps first do some method-specific stuff, then
	# call the default method for this here:
	NextMethod()
}

estimateParameters.idw = function(object, ...) {

  params = getIntamapParams(object$params, ...)
  idpRange = params$idpRange
  if (is.null(idpRange)) idpRange = seq(0.1, 2.9, 0.1)
  nfold = params$nfold
  if (is.null(nfold)) nfold = 5
	# add parameter estimate
	mse = rep(NA, length(idpRange))
  if ("formulaString" %in% names(object)) formulaString = object$formulaString else formulaString = as.formula("value ~ 1")
	params = getIntamapParams(object$params, ...)
	nmax = params$nmax
	nmin = params$nmin
	omax = params$omax
	maxdist = params$maxdist
	debug.level = params$debug.level
	
 	for (i in seq(along = idpRange)) {
  	mse[i] = mean(krige.cv(formulaString, object$observations, nfold = nfold, 
  	nmax = nmax, nmin = nmin, omax = omax, maxdist = maxdist, set = list(idp = idpRange[i]), verbose = params$debug.level)$residual ** 2)	
	}
  best = which(mse == min(mse))[1]
	object$inverseDistancePower = idpRange[best]
	print(paste("best idp value found is", object$inverseDistancePower, "rmse", sqrt(mse[best])))
	return(object)
}

spatialPredict.idw = function(object, ...) {

  
  params = getIntamapParams(object$params, ...)
  nmax = params$nmax
  nmin = params$nmin
  omax = params$omax
  maxdist = params$maxdist
  debug.level = params$debug.level
  
  if (!all(names(object$outputWhat) == "mean"))
    stop(paste("It is not possible to request other prediction types than mean for method idw",
          "requested",names(object$outputWhat))) 
  if ("formulaString" %in% names(object)) formulaString = object$formulaString else formulaString = as.formula("value ~ 1")
  
  if (is.null(maxdist)) maxdist = Inf
 	object$predictions = idw(formulaString, object$observations, object$predictionLocations, 
		nmax = nmax, nmin = nmin, omax = omax, maxdist = maxdist, idp = object$inverseDistancePower, debug.level = debug.level)
	return(object)
}

postProcess.idw = function(object, ...) {
	# smooth over boundaries?

	# spatial aggregation?

	# find out what to output

	# write to data base
  object$predictions = object$predictions[,-2]
  object = NextMethod(object,...)
	return(object)
}