help
help > RE: Maximum number of permutations?
Jan 16, 2017 11:01 PM | Andrew Zalesky
RE: Maximum number of permutations?
Hi Christian,
You should set the number of permutations to 5000 or 10000. It does not matter if the same permutation is sampled multiple times due to chance.
Even if you set the number of permutations to 3003, it is possible for the same number of permutations to be sampled multiple times.
Given that the sampling is uniform, if you sample enough permutations (i.e. 10000), the p-value with random sampling will converge to the p-value with the 3003 distinct permutations.
Andrew
Originally posted by Christian Clemm:
Dear Andrew,
I have a group comparison with two groups, n=6+8.
According to combinatorics, there are 3003 possible permutations in this case. Should I put this number as the number of permutations? Or are the permutations allowed to repeat anyway, so that I can also put e.g. 5,000 or 10,000?
Thank you so much for your help.
Christian
You should set the number of permutations to 5000 or 10000. It does not matter if the same permutation is sampled multiple times due to chance.
Even if you set the number of permutations to 3003, it is possible for the same number of permutations to be sampled multiple times.
Given that the sampling is uniform, if you sample enough permutations (i.e. 10000), the p-value with random sampling will converge to the p-value with the 3003 distinct permutations.
Andrew
Originally posted by Christian Clemm:
Dear Andrew,
I have a group comparison with two groups, n=6+8.
According to combinatorics, there are 3003 possible permutations in this case. Should I put this number as the number of permutations? Or are the permutations allowed to repeat anyway, so that I can also put e.g. 5,000 or 10,000?
Thank you so much for your help.
Christian
Threaded View
| Title | Author | Date |
|---|---|---|
| Christian Clemm | Jan 16, 2017 | |
| Andrew Zalesky | Jan 16, 2017 | |