分位数回归模型在新疆涂阳肺结核报告率生态学影响因素的应用分析  

Quantile regression analysis of ecological factors influencing the reported rate of smear-positive pulmonary tuberculosis in Xinjiang

作　　者：张伟文[1] 贺湘焱[2] 李奇凤[2] 木哈达斯.吐尔逊依明 曹明芹[1] ZHANG Wei -wen;HE Xiang -yan;LI Qi -feng;MUHADASI Tuerxunyiming;CAO Ming -qin(School of Public Health,Xinjiang Medical University,Urumqi,Xinjiang 830011,China)

机构地区：[1]新疆医科大学公共卫生学院,新疆乌鲁木齐830011 [2]新疆维吾尔自治区人民医院,新疆乌鲁木齐830001

出　　处：《现代预防医学》2019年第3期385-388,393,共5页Modern Preventive Medicine

基　　金：2018年新疆维吾尔自治区自然基金项目;No:2018D01C127

摘　　要：目的探讨分位数回归模型在新疆地区涂阳肺结核报告率与生态学相关因素关联分析中的应用价值。方法基于2011-2015年新疆103个区县涂阳肺结核报告率的面板数据,描述与分析其地区分布现况,分析人口密度、人均GDP、年平均温度、年总降水量及年平均相对湿度与涂阳肺结核报告率的关联性,应用分位数回归模型,探讨影响因素随涂阳肺结核报告率变化其效应变化趋势。结果新疆不同地区涂阳肺结核报告率存在地区差异,南疆地区高于北疆、东疆及中部地区,2011-2015年报告率呈小幅波动但整体下降趋势。经典线性模型显示人均GDP越高、年总降水量越高,涂阳肺结核报告率越低,关联有统计学意义(P <0. 05);分位数回归模型进一步提示:随着涂阳肺结核报告率的增加,人均GDP的效应大小呈倒"U"型趋势,年总降水量的影响效应呈递增趋势。结论相比经典线性回归模型,分位数回归模型可提供更加丰富细致的信息,在传染病时空面板数据中具有一定的应用价值。Objective To explore the application value of the quantile regression model in the correlation analysis of the reported rate of smear-positive pulmonary tuberculosis with ecological related factors in Xinjiang. Methods Based on the panel data of the reported rate of smear-positive pulmonary tuberculosis in 103 counties of Xinjiang from 2011 to 2015,the situation of regional distribution and the correlation of the reported rate of smear-positive pulmonary tuberculosis with the population density,per capita GDP,annual average temperature,annual total rainfall and annual average relative humidity were described and analyzed. The quantile regression model was used to explore the trend of the influencing factors with the reported rate of smear-positive pulmonary tuberculosis. Results There were regional differences in the reported rate of smear-positive pulmonary tuberculosis in different regions of Xinjiang. The reported rate of southern was higher than northern,eastern and central in Xinjiang. From 2011 to 2015,the reported rate showed a slight fluctuation,but the overall trend was downward. The classical linear model showed that the higher per capita GDP and total annual rainfall,the lower reported rate of smear-positive pulmonary tuberculosis,and the correlation was statistically significant( P < 0. 05). The quantile regression model further suggested that with the increase of the reported rate of smear-positive pulmonary tuberculosis,the effect of per capita GDP showed an inverted"U"trend,and the effect of annual total rainfall showed an increasing trend. Conclusion Compared with the classical linear regression model,the quantile regression model can provide more detailed information and has application value in the time and space panel data of infectious diseases.

关 键 词：涂阳肺结核 面板数据 分位数回归 

分 类 号：R195[医药卫生—卫生统计学]