PCE with existing data but unknown distribution

UQ with UQLab Community Q&A and How To
,
1

Hello,

I have existing data with an input size of 30 by 7 (7 parameters and 30 subjects), and I have 30 outputs. I want to create Polynomial Chaos Expansions (PCEs) with this data; however, the problem is that, because the data was collected experimentally from a limited population, I do not know the exact details about the distributions of the input parameters.

What is the best way to create a PCE for this? Is it possible to use unknown distributions for creating PCEs in uQLab?

Thanks in advance for your help!

2

Dear @Aep93

UQLab allows you to pass data from any distribution, as explained in section 1.3.2 of the manual:

However, 30 samples for a problem with 7 input variables sounds very ambitious to me. Let’s see how the surrogate performs, but I expect you’ll need a lot more samples unless your problem is really simple.

Best regards
Styfen

1 Like