Statistics Seminars: Algorithms for the Implementation and Evaluation of Monte Carlo Tests
24 October 2011 14:00 in CM221
This talk presents two algorithms concerning Monte Carlo tests (such as a bootstrap or permutation test). The first implements Monte Carlo tests with a uniform bound on the resampling error. The second generates a conservative conﬁdence interval of a speciﬁed length and coverage probability for the power of a Monte Carlo test. These are the first methods that achieve these aims for almost any Monte Carlo test. Previous research has mostly focused on obtaining as accurate a result as possible for a ﬁxed computational eﬀort, without a guaranteed precision. In the proposed algorithms, the computational eﬀort is random and there is a guaranteed precision. For example, the second algorithms operates until a conﬁdence interval can be constructed that meets the requirements of the user, in terms of length and coverage probability.
Contact Ian Jermyn