Statistics Seminars: Non-parametric homogeneity tests with imprecise data
14 December 2011 16:00 in E102
We consider the problem of performing a non-parametric homogeneity test (Kolmogorov-Smirnov or Cramer-Von-Mises) on samples that are imprecisely observed and given as sets of intervals. In this situation, test values become imprecise as well, and the test can result in an additional outcome corresponding to unknown outcome. As computing possible values of the tests is computationally complex, we propose to use the notion of p-box to provide conservative bounds. We study some properties of these approximations, and perform various experiments to study the behaviour of these approximations. Finally, we apply our method to the comparison of medical images whose reconstruction lead to interval-valued images.
Contact email@example.com for more information