流行病学与统计学相结合进行因果探讨的理论方法与研究

In order to study the causality in the epidemiological practice, an advanced study which itegrated modern epidemiological and statistic methods was carried out. (1) We discuss the necessity and sufficiency of the two criteria which provided by Miettinen and Cook, and the concepts of confounders for causal effects at different level are also defined. We also defined the concepts of non-related factors and occasional confounder. (2) We build a new cause-effects model based on the counterfactual model to study the interaction effects. The baseline effects, expose effects and apparent effects are defined in the model，and build a new evaluating system to study the interaction effects base on the non-interaction model. (3) We defined the exchangeable condition and provided a method using co-variances estimating the average causal effects. The methods of confounding and bias controlling were also discussed in our model. (4) The complex graphical model.methods usually are used for the study of complex data analysis and the treating of incompletion data. We divided the large graphical network model into several smaller network and developed new methods of causal inference base on it. Otherwise, the method also resolved the problem of network decomposing with incompletion data, and the partial filling up of EM and MCMC methods was developed which has a faster constringency speed.

..在流行病学因果探索的定性研究基础上，建立横断面与纵向因果推断的定量识别统计理论和方法；结合慢性病分析中流行病的典型问题，在现代生物统计理论与流行病学方法的基础上，进行因果推断的量化识别、理论方法研究，纵向与复杂数据的分析方法上有所突破与创新；开发目前国际上尚未出现的具有因果统计分析方法的流行病学实用软件。..