Hello,
The documentation around the confidence level of Sobol indices is a bit sparse. Could someone explain how the following works in GlobalSensitivity
:
- The role of
nboot
- The role of setting
conf_level
- How to retrieve the confidence level, and the format in which is is returned
Thank you
This doesn’t mention what a bootstrap run does however. Similarly, how does changing the confidence level change anything? Will the number of samples run be changed by the program as a function of this?
Also docs suggest nboot>0
for confidence interval calculation but if nboot=1
then confidence interval objects are nothing
.
The sobol_method.jl was helpful thank you. To clarify my understanding sobol.ST_Conf_Int
will return a matrix where each row is a confidence number and each column is a variable passed? So:
variable 1 variable 2
lower conf level
upper conf level
Bootstrap is Bootstrapping (statistics) - Wikipedia
Similarly, how does changing the confidence level change anything
That changes only the level (pretty well explained in Confidence interval - Wikipedia just in case you are not familiar) and not the number of samples, nboot
changes the number of samples
. To clarify my understanding sobol.ST_Conf_Int
will return a matrix where each row is a confidence number and each column is a variable passed
EDIT: Sorry I misread this before.
No, the way the results are reported for first and total order indices’ CI is
param1 param2 param3 .. ... ...
u1
u2
with each cell being the error for that parameter (the columns)'s sensitivity index for the observed variable (row). So that’s why you have a 4*12 matrix for 4 observed variables and 12 parameters. To get the CIs for the indices you’d do S1[I,j] - S1_Conf_Int[I,j] and S1[I,j] + S1_Conf_Int[I,j] (or through broadcasting)
@Vaibhavdixit02 why does the confidence interval matrix have four rows for each variable? nboot=5
sobol_result.ST_Conf_Int = [
0.002140303660231213 0.001374899485938175 0.0007512253392035508 0.0 0.0 0.0 0.0 0.0 0.0 0.0019364005560527171 0.0015194976539914867 0.0;
3.987682697901888e-7 0.020335209666973064 0.03977866519800592 0.0 0.0 0.0 0.0 0.0 0.0 1.610099093745705e-5 6.116174091283797e-7 0.0;
0.0012173805419186411 0.0015130579222676734 0.004489237720529398 0.0 0.0 0.0 0.0 0.0 0.0 0.00015165970084130953 0.001567894681151645 0.0;
0.0020468867516150084 2.2522540774355426e-5 0.003862861549576147 0.0 0.0 0.0 0.0 0.0 0.0 0.0007652690162809823 0.003977774483835621 0.0]
sobol_result.S1_Conf_Int = [
0.0021595131379424856 0.000518177371997443 0.00010912421052301298 0.0 0.0 0.0 0.0 0.0 0.0 0.001138612347448014 0.0002745154899587079 0.0;
1.2733062154870762e-17 0.04511870593914688 1.8830925047775132e-17 0.0 0.0 0.0 0.0 0.0 0.0 3.319205919341837e-22 1.6983075940831996e-16 0.0;
0.0002551911399908023 0.0009281213495466689 0.00013657384425105572 0.0 0.0 0.0 0.0 0.0 0.0 8.425524493092775e-5 0.004062292604195716 0.0;
0.000757001254193339 2.152668319282848e-5 0.0004343449416748147 0.0 0.0 0.0 0.0 0.0 0.0 4.169105270114894e-8 0.004395152000136753 0.0]