哪些研究问题需要用分层线性模型回答

What Kinds of Research Questions Need to Be Answered by Hierarchical Linear Model

采用常规的回归分析或方差分析处理行为科学与社会科学研究中经常遇到的嵌套数据 ,不能满足独立观察的假定 ,得到的标准误差较小 ,导致I型错误扩大化 ;同时 ,也不利于对不同层次变量不同作用的探讨。分层线性模型则明确区分数据层次 ,通过对个体水平变量和组别水平变量的分层综合分析 ,避免了上述弊病 ,因而可对个体水平的变量进行更准确的预测和更合理的解释。通过实例介绍了将研究问题与分层线性模型有机结合的方法。

In Behavioral and social sciences nested data structure has often been analyzed by traditional regression analysis and analysis of variance. However, nested data structure does not satisfy the assumption of independent observation and has too small standard error. Therefore using those traditional techniques to cope with nested data will lead to inflation of Type I error and hinder the exploration of different functions among variables of different levels. Hierarchical linear model (HLM) clearly distinguishes between levels of data and therefore avoid the above problems through hierarchical and integrated analysis of variables at an individual level and group levels. In doing so HLM can achieve better prediction and explanation of variables of the individual level. The method of dynamically integrating research questions and HLM was introduced with empirical research examples.