四种传染病疫情病死率估计模型的比较

发布时间：2018-02-20 04:28

  本文关键词： 病死率 个体型数据 数值模拟　出处：《科学技术与工程》2017年24期 　论文类型：期刊论文

【摘要】：探讨四种病死率估计模型在疫情初期,对不同疫情类型、疫情规模下估计效能的差异。根据病死率恒定与逐渐下降两种模拟设计,以样本量为1 500及200分别完成四种估计模型、两个真实估计值在四个时间点的1 000次独立估计。病死率恒定时,Modified Kaplan-Meier法和基于gamma分布的参数混合模型的方法具有明显优势;病死率逐渐降低时,早期Modified Kaplan-Meier法、基于gamma分布的参数混合模型的方法以及治愈—死亡危险率比半参数模型的方法均能较理想估计病死率,后期效果欠佳,且敏感性低;但Modified Kaplan-Meier法和SIR模型与竞争风险模型相结合的估计模型的方法在后期无论在精度还是敏感性都较好。样本量的差异对模型估计效能影响不大。说明Modified Kaplan-Meier法和基于gamma分布的参数混合模型的方法在病死率恒定时以及逐渐递减的早期能较准确估计病死率,在逐渐递减后期Kaplan-Meier法和SIR模型与竞争风险模型相结合的估计模型的方法效果有优势;疫情规模对其影响不大。
[Abstract]:In the initial stage of epidemic situation, four kinds of mortality estimation models were studied. According to the two simulation designs of constant mortality and gradual decline, four estimation models were completed with sample size of 1 500 and 200, respectively, for different epidemic types and epidemic scale. 1 000 independent estimates of the two true estimators at four time points. The modified Kaplan-Meier method and the mixed model based on gamma distribution have obvious advantages when the mortality is constant, and the early Modified Kaplan-Meier method is used when the mortality decreases gradually. The method of parameter mixing model based on gamma distribution and the method of cure death risk rate than half parameter model can estimate the death rate better than that of the half parameter model, the late effect is not good, and the sensitivity is low. However, the Modified Kaplan-Meier method and the SIR model combined with the competitive risk model have better accuracy and sensitivity in the later stage. The difference of sample size has little effect on the estimation effectiveness of the model. The results show that the Modified Kaplan-Meier method and the gamma based method have little effect on the estimation effectiveness of the model. The method of distributed parametric mixed model can estimate the mortality more accurately when the mortality is constant and at the early stage of gradual decline. In the later stage of gradual decline, the Kaplan-Meier method and the SIR model combined with the competitive risk model have advantages in the effectiveness of the method, but the scale of epidemic situation has little effect on it.
【作者单位】： 南方医科大学公共卫生学院生物统计学系;西安文理学院数学系;暨南大学经济学院统计学系;
【基金】：西安市科技计划(CXY1531WL40) 国家自然科学基金(81202288)资助
【分类号】：O212.1;R181.3
，

本文编号：1518763