I am using bootstrap for my simulation.

The number of the population is flexible for each case, and the sample size is decided by a certain percentage. For example, I have a 10,000 population, and I decide to use 10% for each iteration of bootstrap, so the sample size is 1,000.

In practice, I found it is hard to decide how many times to run the bootstrap is enough. With less simulation, the results appear insufficiant, while with a large number of simulation they are purely redundant.

May I know if there is a method that can help me to decide the number of iterations to run?

  • 3
    $\begingroup$ If you already have the population at hand, why would you use bootstrap samples to make inference about this population? $\endgroup$
    – Michael M
    Oct 11, 2013 at 11:14
  • $\begingroup$ @MichaelMayer Yes, I could. The problem is there are many populations, and it could be better if I have an uniform method to do so. :) $\endgroup$ Oct 11, 2013 at 12:35
  • $\begingroup$ I think @MichaelMayer is asking why you are trying to infer a population parameter when you have data from the entire population available, so you don't need to infer, you may simply measure. And my guess is that you did not mean to imply you have the entire population sampled; your 10,000 is actually a sample. $\endgroup$
    – jona
    Oct 14, 2013 at 10:36
  • 1
    $\begingroup$ I simply don't understand the question... $\endgroup$
    – Elvis
    Oct 19, 2013 at 14:28

1 Answer 1


The recommended number of bootstrap replications may vary according to "the test to be run on the data"(Mooney, C. Z., and R. D. Duval. Bootstrapping: A Nonparametric Approach to Statistical Inference. Newbury Park, CA: Sage. 1993:11), For standard errors, the same source recommends 50-200 bootstrap replications.


Not the answer you're looking for? Browse other questions tagged or ask your own question.