摘要
目的 探讨树型分类方法在流行病学队列资料分析中的应用价值。方法 依据人群的同质性将总人群根据某一变量 (在流行病学分析中 ,通常为危险因素 )分成不同的亚人群 (在不同的亚人群中疾病的发生概率不同 )。结果 各危险因素在疾病发生中的作用、作用方式 ;危险因素间的交互作用的估计 ;疾病高危人群的筛选。结论 树型分类对多元分类资料进行分析时 ,不涉及参数推断 ,原理简单 ,尤其适用于有共线性的资料处理。在疾病的危险因素分析、以及变量筛选等方面将具有广泛的应用前景。
Objective To explore the applicability of tree-based calssification in epidemiologic cohort data.Methods With potential risk factors included as independent variables,study cohort was divided into several sub-cohorts,according to their homogeneity(i.e.,lung cancer incidence rates).Results Baseline age,Arsenic cumulative exposure and smoking are important factors while splited study cohort.It is found that there is an interaction effect between baseline age and smoking.Meanwhile,the dose-response effect is not maqnitude for arsenic cumulative exposure dose less than 1 774.62ug/m 3.Conclusion The technique of tree-based classification is useful to explore the effect of risk factors to disesase of interest and interaction among risk factors.In addition,classification tree is useable to screen high risk population.Therefore,it could be broadly applied in the screening of covariates and analysis of risk factors.
出处
《中国卫生统计》
CSCD
北大核心
2000年第5期272-275,共4页
Chinese Journal of Health Statistics
