Generating non-gaussian samples from TENDL
AnderGray opened this issue · 4 comments
Hi,
I'm using sandy to generate some random samples of Fe56 from TENDL 2017, and it's giving me some non-gaussian samples.
Would you have any clues as to why this is happening?
The blue is the mean, with the red being individual samples
The endf file I began this calculation with, if that's useful.
Hi @AnderGray,
It is indeed very strange that SANDY does not produce Gaussian samples. I'm afraid I can't help you on this particular point, as I've never looked closely at SANDY's sources.
However, here's my two cents, if it can help you. You have another option: you can start directly from the random files of TENDL. IMHO, this approach is even better: as far as I understand, TENDL's covariances are computed from its random files, which are not necessarily Gaussian. Sampling in TENDL's covariances will produce a degraded (Gaussianized) version of the initial information.
You will (again!) have non-Gaussian distributions, especially in the fast domain, but this time it will be a desirable feature and not a bug-like behavior.
As far as I know, TENDL's covariances are provided for applications that require covariances, such as perturbation approaches.
Hi @vivian-salino,
Thanks for your comment, and for sure using the random endf files is a better option, since it requires no sampling and you use the original files (non-gaussian distribution). But unfortunately the variety of nuclides which are currently available from the TENDL website is quite limited, so I've had to use SANDY for most of the nuclides.