用多元回归分析小儿手术麻醉期间潮气量影响因素

Diagnosis Test and Ridge Regression for the Effect Factors of Tidal Volume

目的确定小儿手术麻醉期间机械通气时潮气量的影响因素。方法采用ＳＡＳ软件对资料进行残差分析，共线诊断并利用原点迫近回归实现岭回归分析。结果在残差图上，学生化残差大致落在｜ｒ｜≤２的水平带状区域内且无任何趋势。一般回归分析的标准误大多大于１．５；身高和体重的方差构成比分别为９０．８７％和５１．３９％，均超过了５０％；在岭迹图２中，当ｋ＝０．３时，岭迹大体稳定，据此所拟合的岭回归方程的所有回归系数标准误均小于０．１。结论残差分析表明本资料近似满足误差方差齐性假设，模型正态性假设；但由于回归系数标准误偏大，因此利用一般回归分析方法所建立的回归方程系数不够稳定，加之用诊断检验查出身高和体重间存在多重共线性，故本资料用一般回归分析方法拟合并不适合；采用岭估计重新分析，不仅减小了多元共线性的效应，而且降低了回归系数的标准误，获得了较为可靠的结果：确认体重、年龄、平均动脉压对潮气量起重要作用。

Objective To determine factors which have important effects on the tidal volume during anaesthesia preshool children with mechanical ventilation in operation. Methods First we utilize Proc Reg in SAS software which performs residual analysis and collinearity diagnosis. Then we complete 'ridge regression with the help of the option of Restrict and the Proc Standard. Results A scatterplot of studentized residual shows the random and uniform dispersion about the zero line of the residual ; Most of standard errors are larger than 1. 5 in the classical regression analysis ; The variance decomposition proportions of height and weight all exceed 0. 5 ; In ridge regression analysis ,all of standard error is smaller than01. Conclusion Although the analysis of residual plot shows that the data comform to the assumptions of equal variance and normal distribution,we think it improper to apply the classical regression analysis for the data because of the instability of regression coefficients and the collinearity between height and weight ; While ridge estimate eliminates these problems ; We conclude that the most important factors which have effects on tidal volume is WEIGHT , and the next are AGE and mean airway pressure (MAP).