Key Randomized Runs in refers to the number K of randomized independent sobol simulations.
The default is K = 1, which means that only one sobol simulation is run with a sample size N as specified in
It should be set to greater than 1 only if an error estimate of the simulated average is desired.
In that case, the total sample size is still equal to N, but rather than one single sobol simulation, K independent smaller sobol simulations run, each of which contains N/K samples.
Each of these sobol simulations produces a partial average output value for the simulated quantity.
The final result over all the simulations equals the average of the thus produced K partial averages.
The estimated error of the final result is then computed in the standard fashion using the sum of the squared deviations of these values from their average.
If = Low Discrepancy, the number here should optimally be set as described at
If the entry implies parallel execution, the number here must be optimally be an exact multiple of the available number of logical processing cores in the device where the simulation runs.