序列累计和数据Joinpoint回归模型构建及应用研究  

Construction and application of joinpoint regression model for series cumulative data

作　　者：曾四清[1] Zeng Siqing(Guangdong Provincial Institute of Public Health/Guangdong Provincial Center for Disease Control and Prevention,Guangzhou 511430,China)

机构地区：[1]广东省疾病预防控制中心广东省公共卫生研究院,广州511430

出　　处：《中华预防医学杂志》2019年第10期1075-1080,共6页Chinese Journal of Preventive Medicine

摘　　要：本研究根据Joinpoint回归模型原理和Poisson分布的可加性,构建序列累计和数据Joinpoint回归模型,以广东省2008—2017年登革热周发病数、周累计和发病数为例进行分析,并以均方差(MSE)和平均相对误差绝对值(MAPE)为指标评估模型拟合效果。除2015年外,其他年份基于周累计和发病数的序列累计和数据对数线性Joinpoint回归模型的MSE和MAPE值均小于基于周发病数的对数线性模型。序列累计和数据Joinpoint回归模型拟合精确度较好,可适用于传染病流行趋势变化特征分析和阶段性累计发病数预测。Based on the principle of Joinpoint regression(JPR)model and the additivity of Poisson distribution,this paper constructed a JPR model for series cumulative data.The notifiable incidence number of dengue fever cases per week and weekly cumulative data in Guangdong province from 2008 to 2017 were analyzed,using(mean squared errors)MSE and(mean absolute percentage error)MAPE to evaluate different models.Except for 2015,the MSE and MAPE produced from the logarithmic linear JPR model based on weekly cumulative incidence number were smaller than those based on the weekly data.The fitting accuracy of JPR model for series cumulative data for trend analysis had been improved significantly.This model could be applied to the analysis of the trend change and the prediction of staged cumulative incidence.

关 键 词：模型 统计学 回归分析 序列累计和数据 

分 类 号：R51[医药卫生—内科学]