Hormesis often refers to a non-monotonic dose-response relationship with beneficial effects at low doses and toxic effects at high doses. It is sometimes referred to as a J- or U-shaped dose-response curve. … There has been long debate whether hormesis theory can be accepted for protecting public health, and this presentation is not to argue whether hormesis theory is applicable in practice. Instead, the focus of the presentation is on statistical modeling as some researchers have pointed out lacking formality in (statistical) hypothesis testing procedures in hormesis studies. In practice, researchers often have a small sample size due to logistics and ethics in animal- or human-based experiments. In this case, statisticians specify some mathematical structures to make assumptions about the unknown true dose-response relationship. If simple assumptions describe the truth closely, we can increase statistical power (the probability of concluding hormesis if it exists) without inflating the false positive rate (the probability of concluding hormesis if it does not exist). If the assumptions are too strong and incorrect, the statistical operating characteristics become implausible. It is a statistical and practical challenge because collecting large data is not always feasible, and it is difficult to make strong assumptions before observing data. Furthermore, in an extreme case, two statistical models may lead to different conclusions on the same data. In this presentation, we discuss how different statistical models and experimental designs perform for detecting hormesis.