Can statistical modeling approaches be used to provide support for multicore timing measurements?

In general, statistical modeling approaches such as Queueing Theory are not applicable to timing analysis of multicore software as software timing behavior does not fit standard statistical assumptions. While the analysis of multicore timing results is based on a representation of the measured real distribution of execution times, it would not be correct to try to estimate averages and standard distributions from the data because we cannot assume that it follows any standard statistical distribution without experimental evidence to show this. 

In some cases, software timing behavior may fit standard statistical assumptions, but this is the exception rather than the rule and must be proven before relying on results from statistical modeling.