countmemaybe More stuff on calculating theoretical error bounds

Move that to its own module. Move from newton to brentq so that it works. Add method to find k for a desired value of epsilon.

Sep 25 '13 18:09 danielhfrank

This looks awesome! Maybe we should use find_k so that a user can either specify k or an error bounds when initializing a KMV... thoughts?

Oct 01 '13 19:10 mynameisfiber

Also, find_k should also accept in an expected set size D if it's known... in this case, finding the optimal k is analytic -- k = 2 * (1 / (p^2 * pi) - 1)

Oct 01 '13 19:10 mynameisfiber

Yeah, those are both very reasonable suggestions. Perhaps I'll get around to adding them soon..

On Tue, Oct 1, 2013 at 3:27 PM, Micha Gorelick [email protected] wrote:

Also, find_k should also accept in an expected set size D if it's known... in this case, finding the optimal k is analytic -- k = 2 * (1 / (p^2 * pi) - 1)

Reply to this email directly or view it on GitHub: https://github.com/mynameisfiber/countmemaybe/pull/1#issuecomment-25480033

Oct 01 '13 22:10 danielhfrank

More stuff on calculating theoretical error bounds

Also, find_k should also accept in an expected set size D if it's known... in this case, finding the optimal k is analytic -- k = 2 * (1 / (p^2 * pi) - 1)

Also, `find_k` should also accept in an expected set size `D` if it's known... in this case, finding the optimal `k` is analytic -- `k = 2 * (1 / (p^2 * pi) - 1)`