平滑曲线拟合 - EmpowerStats Data Analysis for.docx

《平滑曲线拟合 - EmpowerStats Data Analysis for.docx》由会员上传分享，免费在线阅读，更多相关内容在教育资源-天天文库。

1、平滑曲线拟合本模块使用广义相加模型(GAM)检验结果变量和危险因素（暴露）之间的非线性关系。它有助于发现非直线性的关系，并判断有无阈值效应。确定用一般的直线性回归是否适当。结果变量分布类型可以是多种，如正态分布、二项分布、泊松分布等，还可以是时间依赖的生存状态（此时使用的RCS:restrictedcubicsmoothing）。若要更改曲线光滑性（自由度），右键单击风险因子(X)，然后在弹出的菜单中单击"更改的自由度"。默认值（设置df=0)表示由最小GCV方法确定的自由度。如果要用平滑线散点图，请选中"散点图绘制原始点"。例：D

2、EMO数据曲线拟合BMI与SBP、DBP的关系。输出结果：GeneralizeadditivemodelsOutcome:SystolicBP,mmhgExposure:Bodymassindex,kg/m2LineartermseffectEstimateStd.ErrortvaluePr(>

3、t

4、)95%CIlow95%CIupp(Intercept)110.07764.28425.69530101.681118.4741factor(SEX)2-5.58162.1559-2.5890.0098-9.8072-1.356AGE0

5、.80810.060213.430600.69020.926OCCU.NEW-4.31331.4716-2.93090.0035-7.1977-1.4289factor(EDU.NEW)20.31521.92950.16340.8703-3.46664.097factor(EDU.NEW)30.45432.30860.19680.8441-4.07064.9792SMOKE-3.6311.9788-1.8350.0669-7.50940.2473ALH-1.73642.1874-0.79380.4276-6.02362.5509Ch

6、i-squaretestsforlineartermsdfFp-valuefactor(SEX)16.70280.0098AGE1180.38030OCCU.NEW18.59040.0035factor(EDU.NEW)20.0210.9792SMOKE13.36730.0669ALH10.63010.4276ApproximatesignificanceofsmoothtermsedfRef.dfFp-values(BMI):factor(SEX)11.3851.6880.10930.8449s(BMI):factor(SEX)2

7、2.18312.81534.4480.0047ModelstatisticsN:784Adj.r-square:0.24Devianceexplained:0.2503UBREscore(sp.criterion):391.8707Scaleestimate:386.088594182483family:gaussianlinkfunction:identityOutcome:DiastolicBP,mmhgExposure:Bodymassindex,kg/m2LineartermseffectEstimateStd.Errort

8、valuePr(>

9、t

10、)95%CIlow95%CIupp(Intercept)57.79762.315624.9601053.25962.3361factor(SEX)21.09191.16440.93780.3487-1.19033.3742AGE0.30050.03269.228200.23670.3644OCCU.NEW-1.03040.7944-1.29710.195-2.58740.5266factor(EDU.NEW)21.72721.04151.65840.0976-0.31413.7685factor(EDU.NE

11、W)32.16561.24661.73710.0828-0.27794.609SMOKE-0.1381.0687-0.12910.8973-2.23261.9566ALH0.44691.18150.37830.7053-1.86882.7627Chi-squaretestsforlineartermsdfFp-valuefactor(SEX)10.87940.3487AGE185.15990OCCU.NEW11.68250.195factor(EDU.NEW)21.80840.1646SMOKE10.01670.8973ALH10.

12、14310.7053ApproximatesignificanceofsmoothtermsedfRef.dfFp-values(BMI):factor(SEX)12.41653.08520.94920.4141s(BMI):fact