Why do the results returned by calling the "simulate" function with a "creditDefaultCopula" object vary based on the order of the rows in the input data?

When using "creditDefaultCopula" from Risk Management Toolbox, why are simulation results different when the rows of the data input are in different order even with the same random seed for all simulations?

Answer 1

[MathWorks Support Team]

This is an expected behavior. The "creditDefaultCopula" object simulates the independent random numbers for all rows at the same time before passing them into the model. Thus, changing the row ordering of the data results in the corresponding independent random number for each row to be different.

However, the simulation results should be comparable and the relative difference between the simulation results should be small. Note that these kinds of credit simulations are slow to converge, and the functionality can handle large number of scenarios. The "creditDefaultCopula" object is expected to support very large number of simulations without running out of memory even on a single machine, and parallel computing is also supported. With more simulations, the results would be more stable, though still not exactly the same if the order of inputs change.