连续型显变量调节效应建模方法的规范与例解

Normalization of moderation effect modeling for continuous manifest variables with an illustrative case

为规范连续型显变量的调节效应建模方法。通过1个例解,比较了原始化、仅对自变量和调节变量进行均值中心化、对所有变量进行均值中心化和对所有变量进行标准化4种建模方法的表面差异和实质内涵。总体而言,尽管原始化方法可能存在着某些潜在的误导之处,但在实质上与均值中心化、标准化方法得到了相同的统计结果。不同建模方法下的回归方程显示的是,在调节变量处于不同水平时,自变量对于因变量的影响程度之差异。而这正是调节效应模型的实质内涵。在课堂教学与学术研究中使用调节效应模型时,应该注重建模方法的规范性并作出恰当的统计解读,其中,标准化处理应该得到更多的应用。

To normalize the moderating effect modeling for continuous manifest variables.With an illustrative case,we compared the apparent differences and substantive connotations across four modeling approaches:primitive model,mean-centering for independent variables and moderating variables alone,mean-centering of all variables,and standardization of all variables.Overall,despite certain pitfalls that could be potentially misleading,the primitive approach yielded essentially the same statistical outcomes as with the mean-centering or the standardization approach.Regression equations derived from these different modeling approaches illustrated varied influence of the independent variables on the dependent variables according to different levels of the moderator,and this,exactly represented the substantive connotations of moderating effect modeling.The use of moderating effect models in classroom practices and academic research should command attentions to normalization of modeling approach and proper interpretation of statistical outcome.Among others,the standardization modeling approach warrants widely use in more settings.