问题
I am working on a model in netlogo which is a statistical physics based model. How many times should i run the simulation and average the results to conclude the study?
回答1:
As Marzy points out in comments, there is no definitive answer. Many people will pick the magic numbers 20-30 out of a hat and go with that. A common choice is to do a small number (5-10) runs initially, estimate the variance from them, and then project how large a sample size you would need to bring the margin of error (half-width of the confidence interval) down to some desired tolerance.
来源：https://stackoverflow.com/questions/32222777/how-many-times-should-i-run-computer-simulations-to-conclude-one-experiment-in-n