Combination of SAMPLER simulations

[Jake Smith]

Hello, 

I am curious if it is possible to combine the outputs from multiple SAMPLER simulations to create a distribution of a post-processed result. Is the process for perturbing nuclear data (XS, yield data, decay data) random so that there would be no correlations in the perturbed cases? For example, if I were to run 50 cases with one SAMPLER execution and then run another 50 cases with another SAMPLER execution, would it be valid to construct a distribution from all 100 cases? Thank you!

[Ugur]

Jake sorry for the double email. Apparently we just replied to you not to the group. So, I will repost here for the records.

By design Sampler produces the same results for every N samples when you don't change your starting sample number. Unless you need more than 1000 samples, you can run your 50 and if you are not happy with the statistics, you can increase your number of samples to 100 and Sampler will only add 50 more samples/runs to the previous one and calculate the statistics of the 100 samples. You can keep this process  until you reach 1000 samples. Or simply create all of your samples at once and ask Sampler not to run any case. You can later submit them in batches as you like once you get all your runs completed, you can run Sampler and it will process them as they all came from the same distribution.

thanks
Ugur Merturek