<正>In this paper, we consider the risk assessment problem under multi-levels and multiple mixture subpopulations. Our result is the generalization of the results of [1-5].1 Finite Mixture Normal ModelsIn dose-response studies, a class of phenomena that frequently occur are that experimental subjects (e.g., mice) may have different responses like 'none, mild, severe' after a toxicant experiment, or 'getting worse, no change, getting better' after a medical treatment, etc. These phenomena have attracted the attention of many researchers in recent years. Finite