For small vectors, this can give us a big computational speedup. There are (2n-1) choose n possible resamples of a vector. When this value is smaller than the number of bootstraps, we should do the exact test instead.
We talked about this a while back (to say the least!) -- our caching approach will limit the impact of this optimization, but of course, the exact test makes sense anyways.
For small vectors, this can give us a big computational speedup. There are (2n-1) choose n possible resamples of a vector. When this value is smaller than the number of bootstraps, we should do the exact test instead.
http://www.wolframalpha.com/input/?i=%282n-1%29+choose+n