sensitivity (version 1.7)

sobolCert: Monte Carlo Estimation of Sobol' Indices using certified meta-models

Description

sobolCert implements the Monte Carlo estimation of the Sobol' sensitivity indices using certified metamodels using the formulas in Janon et al. (2011).

Usage

sobolCert(model = NULL, X1=NULL, X2=NULL, nboot = 300, conf = 0.95, lambda0 = 0, h = 0)
## S3 method for class 'sobolCert':
print(x, \dots)

Arguments

model

a function defining the model to analyze. This function must return a list whose components are:

• out

{metamodel output.} err{metamodel output error bound, satisfying $$|model_{output} - metamodel_{output}|

Value

• sobolCert returns a list of class "sobolCert", containing the following components:
• call
  the matched call.
• S
  the estimations of the Sobol' sensitivity indices.
• penalty
  (only if lambda0>0) value of the smoothing penalty.

item

• X1
• X2
• nboot
• conf
• lambda0
• h
• x
• ...

References

Janon, A., Nodet M., Prieur C. (2011) Uncertainties assessment in global sensitivity indices estimation from metamodels.

See Also

sobol, sobol2002, sobol2007

Examples

# Test case

n <- 1000
X1 <- data.frame(matrix(runif(3 * n), nrow = n))
X2 <- data.frame(matrix(runif(3 * n), nrow = n))

# sensitivity analysis
x=sobolCert(model=function(X) { list(out=X[1]+2*X[2]+X[3]+.001*runif(1),err=.01); }, 
            X1, X2, conf=.99, lambda0=.1, h=.1, nboot=30)
print(x)

x=sobolCert(model=NULL, X1=NULL, X2=NULL, conf=.95)
print(x)