限制性立方样条在非线性回归中的应用研究

The Application of Restricted Cubic Spline in Nonlinear Regression

目的探讨限制性立方样条在自变量和应变量非线性相关时的应用。方法应用限制性立方样条分析泰坦尼克号乘客生存数据,通过LOWESS平滑曲线和AIC值比较限制性立方样条和分段亚元回归的分析效果。结果限制性立方样条结合logistic回归可以较好地拟合自变量和应变量的非线性关系,并能发现重要的关键点。结论限制性立方样条在非线性影响因素分析,特别是剂量效应分析中有比较好的应用前景。

Objective To explore the application of restricted cubic spline（RCS） in nonlinear regression.Methods Restricted cubic spline was used to analyze the survival data of Titanic passengers.LOWESS curve and AIC value were used to compare RCS with categorized regression.Results RCS combined with logistic regression fitted nonlinear regression well and also was able to find out key points.Conclusion RCS should be used more extensively in nonlinear regression in causal inference and dose response model.