KRIG (version 0.1.0)

Krigvar: Combinatorial variance computation.

Description

Computation of variance

Usage

Krigvar(KF, Gamma)

Arguments

KF

values of the kernel integral evaluations.

Gamma

Cube with integral results.

Value

Real value of sensitivity.

References

Kanova:2013KRIG Aronszajn:ThRKKRIG

Examples

# NOT RUN {
library( KRIG ) 
options( stringsAsFactors = FALSE )
 
kernel_1<-function( x, y ) exp( -0.5*(x-y)^2)
kernel_2<-function( x, y ) exp( -0.7*(x-y)^2)
kernel_3<-function( x, y ) exp( -0.1*(x-y)^2)

Kernels<-data.frame( kernel = c( 'kernel_1', 'kernel_2', 'kernel_3' ), 
                     min = c( -1, -1, -2 ), 
                     max = c( 1, 1, 2 ),
                     n = c( 100, 100, 100 ) )
                     
n<-20
X<-matrix( c( seq( -1, 1, length.out = n ), 
              seq( -1, 1, length.out = n ),
              seq( -2, 2, length.out = n ) ), n, 3 )
              
KI<-list_integrate_kernel( Kernels, X )
GK<-Kanova( Kernels, KI, X )
    
f<-function( x ) x[1] + 30 * x[2] + 60 * x[3]
Func<-apply( X, 1, FUN = f )
    
KF<-solve( GK$Kanova + diag( 1e-8, n, n ), Func )
    
SbI<-NULL
for ( j in 1:3 ) {
  CB<-combn( 1:3, j )  
  for ( l in 1:ncol( CB ) ) {
    SbI<-c( SbI, Krigidx( KF, CB[,l], X, GK$Gamma ) )
    names(SbI)[length(SbI)]<-paste( 'C.', paste( CB[,l], collapse='.' ), sep = '' )
  }
}
  
Var<-Krigvar( KF, GK$Gamma )
    
SVar<-sum( SbI / Var )

# }