vgramfit<-function(ds,nds,evals,vgram,p0){
  #function to fit variogram models using Cressie's
  #weighted least squares criterion
  #
  #ds is a vector of distances or displacements at which
  #empirical variogram values have been computed, nds gives
  #the number of pairs of loctions used for each value,and
  #evals should be a vector of those empirical variogram
  #values
  #
  #vgram is the name of a function giving the theoretical
  #variogram model -- this should take arguments of (1) parameters 
  #needed in the variogram model, and (2) distances
  #or displacements (or midpoints of classes of these) as a 
  #vector. The function vgram should return a vector of evaluated
  #variogram values (semi-variogram)
  #
  #p0 is a vector of starting parameter values
  #
  #define the weighted optimization function
  Qfnct<-function(ps,hs,nhs,ghats){
     gvals<-vgram(ps,hs)
     crit<-sum(nhs*((ghats/gvals)-1)^2)
     return(crit)
     }
  #fit using the generic optimization function in R
  fit1<-optim(p0,Qfnct,hs=ds,nhs=nds,ghats=evals)
  #fit using nonlinear minimization function in R
  #fit2<-nlm(Qfctn,p0,hs=ds,nhs=nds,ghats=evals)
  res1ps<-fit1[[1]]; res1crit<-fit1[[2]]
  #res2ps<-fit2[[2]]; res2crit<-fit2[[1]]
  #res<-list(res1ps,res1crit,res2ps,res2crit)
  res<-list(res1ps,res1crit)
  return(res)
}