Difference Detection Between Two Contrast Sets

摘要

Mining group differences is useful in many applications, such as medical research, social network analysis and link discovery. The differences between groups can be measured from either statistical or data mining perspective. In this paper, we propose an empirical likelihood (EL) based strategy of building confidence intervals for the mean and distribution differences between two contrasting groups. In our approach we take into account the structure (semi-parametric) of groups, and experimentally evaluate the proposed approach using both simulated and real-world data. The results demonstrate that our approach is effective in building confidence intervals for group differences such as mean and distribution function.