现场调查中零频数过多的统计分析方法

A statistical methods to extra zero in the data of field suvery

目的:研究零过多计数资料的分布拟合方法并应用于急性上呼吸道感染调查资料。方法:将数据分为两部分,一部分为零部分,另一部分为Poisson分布部分,然后通过转换,形成条件零堆积模型(ZTP),运用logit和log两种连接函数进行拟合。结果:在ZTP部分表示,低年龄和住在低层的发病次数多;logit部分显示年龄小、有锻炼习惯者、健康状况差和有慢性呼吸系统疾病史的人易发病。结论:转换后的零过多模型可以更好地对影响因素进行解释。

Objective：To study how to model zero-inflated count data, and apply it to handle the data about respiratory infection. Methods：Zero-inflated model with 2 parts,zero and Poisson distribution, were transformed to conditional zero-inflated model. Logit and Log linkage function were used to promote the explanation of the result. Results：The results of significant risk factors were those who were younger and live lower in the ZTP part; the risk factors were younger exerciser, having history of chronic respiratory system disease and ill body conditions in the LOGIT part. Conclusion：The conditional zero-inflated model could be easier to explain the affected factors.