论文部分内容阅读
病例随访研究或队列研究中,当暴露因素和混杂因素间存在交互作用时,用传统的方法估计暴露组的总的相对危险度是无意义的。由于肿瘤学、放射学等随访资料常服从 Poisson 分布,因而可用 Poisson 回归模型拟合并检验混杂因素与交互作用是否存在。当交互作用不存在时,可通过模型参数估计总的相对危险度。本文通过对肺癌不同病程分期资料和不同的随访时间分层进行Poisson 回归模型拟合,其结果表明这种估计是有效的。
In case follow-up studies or cohort studies, when there are interactions between exposure factors and confounding factors, it is pointless to estimate the total relative risk of the exposure group using traditional methods. Since follow-up data such as oncology and radiology often follow the Poisson distribution, a Poisson regression model can be used to fit and test for the existence of confounding factors and interactions. When the interaction does not exist, the total relative risk can be estimated from the model parameters. In this paper, Poisson regression model was used to fit different stages of lung cancer staging data and different follow-up time. The results show that this estimation is effective.