累积残差在广义估计方程模型诊断中的应用

Application of Cumulative Sums of Residuals in GEE Model Checking

目的探讨累积残差在广义估计方程模型诊断中的应用及SAS9.1.3软件实现。方法收集癫痫新疗法临床疗效资料,构建广义估计方程模型,根据QIC准则进行作业相关矩阵选择,采用累积残差检验固定效应及总体均数函数形式是否正确。结果累积残差图直观明确,计算Kolmogorov-type supremum test概率值能客观地检验拟合模型效果,是值得推荐的一种方法。结论对相关数据分析,构建广义估计方程时,采用累积残差技术进行模型诊断,可以获得更为客观的分析结论。

Objective To explore the application of cumulative sums of residuals in GEE model cheching and implement with SAS 9. 1.3 software. Methods Collecting the data of clinical curative effects of new medicine in curing the falling sickness, fitting genrcalized estimating equations （GEE）, selecting the working covariance matrix using Quasilikelihood under the independence model criterion （QIC）, checking the form of the fixed effectives and of the model for the mean by cumulative residuals. Results Sums of residuals figure is clear and identity, Kolmogorov-type supremum test provide objective and informative checks on the adequacy of the fitted model, the technique deserve recommending. Conclusion Suggesting to check the model using cumulative sums of residuals,when fitting the GEE,it can provide more objective results.