Answers are grounded in this paper and cite their sources.
In using a computer experiment to calculate the time correlation function of some dynamical variable, the ensemble average over an equilibrium distribution in phase space often is replaced by a time average over a finite interval T. It is shown here that the statistical uncertainty due to this kind of average is of the order of 1T1/2. The coefficient of the square root is related to a characteristic relaxation time of the correlation function. If only time averaging is performed, the resulting statistical uncertainty in typical computer experiments may be of the order of 20%.
No takes yet. Share an insight, caveat, or question.
Zwanzig et al. (1969) studied this question.
Synapse has enriched 5 related papers. Consider them for comparative context: