Logistic回归模型中自变量相对重要性的优势分析

Dominance Analysis for the Relative Importance of Independent Variable in Logistic Regression Model

目的应用扩展优势分析方法于Logistic回归模型中,为研究者在确定模型中自变量相对重要性提供一种可选择的方法。方法通过计算和比较与某自变量有关的所有可能子模型(即含有该变量的不同组合)的平均贡献增量△R2,以评价该自变量的相对重要性,并应用于实例分析。结果优势分析所得的各变量的总平均贡献之和等于最终模型的决定系数,其重要性排序与标准回归系数的排序不同,且R2M和R2E更适合作为优势分析的指标。结论优势分析可将各自变量对因变量总方差的贡献,分解为已解释方差百分比,且独立于模型,能精确地衡量自变量的相对重要性。

Objective To provide an alternative for researchers in determining the relative importance of independent variables in models by applying extended dominance analysis in logistic regression model. Methods To evaluate the relative importance of independent variable and apply to a ease analysis by calculating and comparing the average incremental contribution AR2 of a variable relative to all possible subset models （all different combinations containing the given variable）. Resdts Dominance analysis results indicated that the overall average contribution of each variable was equal to the R2 of the final model, and the ordering of importance was different from the standardized regression coefficients. Conclusion Dominance analysis can decompose each variable accounting for the total variance into the percentage of explained variance. Furthermore, the results are independent of the models and they measure the relative importance of the independent variables accurately.