• About DI
• Home
• Installation
• Manual/Tutorial   
• Example Dataset
• Related Software
• Fields
• Funfits

DI   offnw.prob

Calculate off-network probabilities.
 	
	
DESCRIPTION:

	Takes a network object and then uses Krig.nw (which in turn 
        uses the Fields function, Krig, to calculate the probability 
        of exceeding a given value on any grid point off of the 
        network. This function is called by plot.prob

USAGE:
	offnw.prob(nw.obj=new.network, Y, K=85, GCV=T, deg.f=NA, 
                   Rho=NA, Sigma2=NA, plts="Probabilities", ...)


REQUIRED ARGUMENTS:

nw.obj
	Network object created by make.network.obj

Y
	Vector of measurements recorded at each point on the network.

K
	Exceedence value; P[ Z > K].  Default is 85.

OPTIONAL ARGUMENTS:

GCV
	Use Generalized Cross-Validation criterion for kriging?

deg.f
	degrees of freedom; for use with the Krig function.

Rho
	correlation parameter for use with the Krig function.

Sigma2
	variance parameter for use with the Krig function.

plts
	Which plot you wish to view.  Choices are: probabilities, 
        the means as found by kriging, the standard deviations as 
        found by kriging, diagnostics of the Krig function or 
	all of the above.  This is more important for the function 
	plot.prob which actually calls offnw.prob


VALUE:
	Returns a list object giving the means, standard deviations, 
	probabilities, etc...

SEE ALSO:
	Krig, Krig.nw, plot.prob

EXAMPLES:
# The function is currently defined as
function(nw.obj = new.network, Y, K = 85, GCV = T, deg.f = NA, 
     Rho = NA, Sigma2 = NA, plts = "Probabilities", ...){
x0 <- make.surface.grid(nw.obj$grid)
y0 <- Y
if(is.null(nw.obj$cov.obj$mean.obj)) {
	mean.obj <- 0
	class(mean.obj) <- "constant"
}
else if(is.numeric(nw.obj$cov.obj$mean.obj)) {
	mean.obj <- nw.obj$cov.obj$mean.obj
	class(mean.obj) <- "constant"
}
else mean.obj <- get(nw.obj$cov.obj$mean.obj)
if(is.null(nw.obj$cov.obj$sd.obj)) {
	sd.obj <- 1
	class(sd.obj) <- "constant"
}
else if(is.numeric(nw.obj$cov.obj$sd.obj)) {
	sd.obj <- nw.obj$cov.obj$sd.obj
	class(sd.obj) <- "constant"
}
else sd.obj <- get(nw.obj$cov.obj$sd.obj)
if(GCV)
	Krig.obj <- Krig.nw(nw.obj, Y = y0, rho = Rho, sigma2 = Sigma2, 
		meanobj = mean.obj, sdobj = sd.obj, ...)
else Krig.obj <- Krig.nw(nw.obj, Y = y0, df = deg.f, rho = Rho, sigma2
		 = Sigma2, meanobj = mean.obj, sdobj = sd.obj, ...)
mtemp <- predict(Krig.obj, x0)
setemp <- predict.se(Krig.obj, x0)
p <- 1 - pnorm((K - mtemp)/setemp)
if(plts == "diagnostics" | plts == "all") {
	list(prob = as.surface(x0, p), mean = as.surface(x0, mtemp),
	se = as.surface(x0, setemp), K = K, fit = Krig.obj)
} else {
	list(prob = as.surface(x0, p), mean = as.surface(x0, mtemp), 
	se = as.surface(x0, setemp), K = K)
}
}