It seams that fast99 function (sensitivity package) does not work out for norm distribution.

The author has already gave the illustration and solution as well.

Thanks.



2013/6/9 Uwe Ligges <ligges at statistik.tu-dortmund.de>

On 08.06.2013 00:54, Marino David wrote:

Dear all mailing listers,

Does Anyone have the same problem as mine when using the fast99
(extended-FAST method) to perform SA of model with norm distribution
inputs?

Why not ask the authors of the function?

Best,
Uwe Ligges

See the simple example given following.

Any suggestion will be greatly appreciated.

Thank you!

Marino


# Simple example

# 1. uniform version (It works well)
library(sensitivity)
Myfun<-function(x){return(**rowSums(x))}
SA1 <- fast99(model = Myfun, factors = 3, n = 1000, q = "qunif", q.arg =
list(min = 0, max = 1))


SA1

Call:
fast99(model = Myfun, factors = 3, n = 1000, q = "qunif", q.arg =
list(min = 0,     max = 1))

Model runs: 3000

Estimations of the indices:
   first order total order
X1   0.3335243   0.3350584
X2   0.3335243   0.3350584
X3   0.3335243   0.3350584






# 2. norm version (it does not work)
SA2 <- fast99(model = Myfun, factors = 3, n = 1000, q = "qnorm", q.arg =
list(mean = 0, sd = 1))


SA2

Call:
fast99(model = Myfun, factors = 3, n = 1000, q = "qnorm", q.arg =
list(mean = 0,     sd = 1))

Model runs: 3000

Estimations of the indices:
   first order total order
X1          NA          NA
X2          NA          NA
X3          NA          NA

       [[alternative HTML version deleted]]