safi (version 1.0)

safiModel: Functional sensitivity analysis

Description

computes normalized regression indices for the sensitivity analysis of functional inputs

Usage

safiModel(s.d, y)

Arguments

s.d

safidesign-object

y

model response

Value

safimodel object containing the design and the computed coefficients

Details

If the design was created with method "SB" the coefficients are computed via sequential bifurcation, for method "other" via least squares estimation.

References

Fruth, J.; Roustant, O.; Kuhnt, S. (2014) Sequential designs for sensitivity analysis of functional inputs in computer experiments, Reliability Engineering & System Safety, doi: 10.1016/j.ress.2014.07.018, preprint on HAL: http://hal.archives-ouvertes.fr/hal-00943509.

Examples

### simple example

s.d <- createSafiDesign(d.f = 1)
s.d2 <- splitSafiDesign(s.d = s.d, new.split.points = list(c(0.25, 0.75)))

# artificial model output (rising influence)
x <- accessSafiDesign(s.d = s.d2, n.timepoints = 4)
y <- x$x1 %*% c(0, 1, 2, 3)
s.m <- safiModel(s.d = s.d2, y = y)
plot(s.m)


### d.f = 3, mirrored

s.d <- createSafiDesign(d.f = 3, mirrored.runs.included = TRUE)
s.d2 <- splitSafiDesign(s.d, list(c(0.5), c(0.25, 0.75), c(0.25, 0.5, 0.75)))

# artificial model output (x1 without influence, x2 rising, x3 falling)
x <- accessSafiDesign(s.d = s.d2, n.timepoints = 4)
y <- x$x2 %*% c(0, 1, 2, 3) + x$x3 %*% c(0, -1, -2, -3)
s.m <- safiModel(s.d2, y = y)
plot(s.m)