Jonathan Feinberg

Depends a bit on what you want to do, but most likely sure. E.g. you can use ``chaospy.SampleDist`` to intepret data as a distribution, and off you go. I'll note...

I doubt you will loose convergence all together, just the fast speed of the convergence PCE usually offers. But I will note that I am not an expert on data...

The parameters `mu` and `sigma` refers to the mean and SD of the cooresponding Gaussian distribution (before the log-transformation). When you set `mu=500` and `sigma=270` you get `È[X] ~= e^125000`...

Then you need to use the method of moments. For log-normal that is a bit involved, but it is described here: https://www.real-statistics.com/?p=30118 Plug in `\bar x=500` and `s=270` and calculated...

PCE with log-normal is notouriously know for being problematic, both analytically and in practice. Dakota defines it as "transformed Normal", which I think means they are doing generalized-PCE. This is...

Thank you @regislebrun, much appriciated. @ambehnam, I've med a new tutorial here: https://chaospy.readthedocs.io/en/master/tutorials/stochastic_dependencies.html not coined towards your problem in particular, as I use dependent random variables as the focus, but...

Yes, that seems correct, though you are outside the edge of what chaospy has been tested for (g-pce+log-normal+high-dim+genz-keister+sparse). I just ran the code on an synthetic example, and it look...

As for your follow-up question: what do you mean "very large numbers?" Like your first problem by going beyond numpy max limit of 1e308? And quadrature order are typically not...

I found the sentence in the original paper: "In general, we found that we could not obtain sequences with more than 5-7 steps, and in any event were limited by...

That would be appriciated if you post the results here so I know what chaospy is capable of. Variance of 1370 is not an issue in and of itself, but...