Hi again,
To clarify my question, what I want is to produce a conditional simulation of my process without the nugget effect. If it is not possible via simtub but the option is available in kriging (basically I just want to remove the Nugget effect that is always the first component of the variogram in my setup), I was thinking to do the conditional simulation in a more indirect way by doing something like:
- Code: Select all
# 0) model without the nugget effect (i.e removing the first basic structure)
data.model2=new('model',ndim=2,nvar=1,basics=data.model@basics[seq(2,length(data.model@basics))])
# 1) non conditional simulations on x_alpha's and x_0's
grid.db=simtub(,grid.db,data.model2,nbsimu=1,nbtuba=1000)
data.db=simtub(,data.db,data.model2,nbsimu=1,nbtuba=1000)
# 2) kriging of the true data on the grid x_0's
data.db=db.locate(data.db,"Gaussian.thick",'z')
grid.db=kriging(data.db,grid.db,data.model,data.u.neigh,cov.extract=c(2:length(data.model@basics)),radix="KS.Part")
# 3) kriging of the non conditional simulated data on the grid x_0's
data.db=db.locate(data.db,"Simu.V1.S1",'z')
grid.db=kriging(data.db,grid.db,data.model,data.u.neigh,cov.extract=c(2:length(data.model@basics)),radix="KS.Part")
# 4) final value of the conditional simulations on x_0's
grid.db=db.add(grid.db,Simu.Gaussian.thick.S1=db.extract(grid.db,"KS.Part.Gaussian.thick.estim")+(db.extract(grid.db,"Simu.V1.S1")-db.extract(grid.db,"KS.Part.Simu.V1.S1.estim")))
Do you think it is correct? Even if it is the case, I am still open to suggestions for having a direct solution with simtub.
Thanks for your help,
Maxime