Jonathan Feinberg

That is an excellent question. I usually don't have no answer for questions like these. My best educated guess is that the problem you are describing happens to be a...

Yes, I figured out that it was problem specific, and that increasing order helps. But I am still curious as to why the convergence rate drops so much as it...

Problem discussed in #188.

1) Are you increaing the polynomial order without increasing the number of evaluated samples? For regression, try to keep the number of evaluated samples roughtly 2 times the number of...

1. yes, that is what I meant.

I am unsure. Could you post the following: * OS or Distro * Version of Python * Version of Chaospy and Numpoly From your example, do mind posting the output...

Ah, the variable names are not sorted for more than 10 variables. That makes sense. I've made a new release of Chaospy now version 4.3.7. which has a fix for...

In principle yes as Sens_m is using E_cond underneath the hood. But I don't think so from skimming the code. Are you getting different results with Sens_m in 4.3.7?

Chaospy is inherently positional. If you rearange the order of something, that is fine, you just need to be consistent about it. The following two methods should be equivalent for...

If you want to keep the marginsl distributions, but induce correlation, I'd look into Copula theory. If you want a simple suggestion, I'd go for something like cp.Nataf and provide...