线性模型中自变量相对重要性优势分析法估计及其应用

Estimation and Application of Dominance Analytic Method for Relative Importance of the Independent Variables in Linear Regression

目的在多元线性回归模型中,估计各自变量的相对重要性,并探索区间估计方法。方法在自变量间存在相关时,运用Budescu(1993),Azen(2003)提出的优势分析法估计肝手术病例预计存活时间的影响因素重要性,并运用Bootstrap法探索区间估计方法以此来评价估计结果的变异性。结果血凝素、预后指数、酶功能对预计存活时间的相对贡献分别为0.1415、0.3408和0.490,其Bootstrap法95%可信区间分别为(0.0573,0.2744)、(0.2359,0.4545)和(0.3411,0.6090)。结论酶功能对肝手术病例预计存活时间的影响最大,预后指数次之,血凝素最小。当自变量间存在相关时,优势分析法估计的自变量相对重要性结果更精确稳定,值得推广应用。

Objective To estimate the degree of importance of each variable in multiple linear model, and explore the methods of confidence interval. Methods When predictor variables were correlated, the method of dominance analysis was used to estimate the importance of influencing factors of survival time d the liver surgery patients, and the bootstrap is u^d to explore the methods of confidence interval and assess the sts.ble of dominance results. Results The relative contribution of hemagglutinin, prognostic index, enzyme function were 0. 1415, 0. 3408 and 0. 490 ,respectively, with bootstrap CIs of （ 0. 0573, 0. 2744 ）, （ 0. 2359, 0. 4545 ） and （ 0. 3411, 0. 6090） respectively. Conclusion The maximum influencing factor is enzyme function, prognostic index, followed by the minimum hemagglutinin. When predictor variables are correlated, the results estimated by dominance analysis are more accurate and stable.