广义可加模型共曲线性及其在空气污染问题研究中的应用

Concurvity in generalized additive models in study of air pollution

目的:针对大气污染与健康关系研究中拟合广义 可加模型时的共曲线性现象进行研究并试图减小其影响. 方法:采用非参数条件自助抽样法来估计模型系数及其可信 区间.结果:实例数据分析可见自助法得到的结果较好. 结论:非参数条件自助抽样法是克服广义可加模型中共曲线 性影响的有益尝试.

AIM: To explore the effect of concurvity of generalized additive models (GAMs) in the study of air pollution and human health and to diminish the influence of concurvity. METHODS: We employed the nonparametric conditional bootstrap method to diminish the influence of concurvity. RESULTS: This method got a better confidence interval. CONCLUSION: The nonparametric conditional bootstrap method can overcome the effect of concurvity in GAMs and get a better result.