我国近十年来心理学研究中HLM方法的应用述评

Review of the Application of Hierarchical Linear Modeling in Psychological Research from 2002 to 2011 in China

以2002-2011年中国期刊网收录的50例应用多层线性模型(HLM)的心理学期刊论文为研究对象,从样本描述、模型发展与规范、数据准备、估计方法与假设检验4个角度进行文献计量和内容分析,对我国心理学研究中HLM方法的使用现状进行评估。结果表明,HLM方法主要用于管理、发展和教育心理学,绝大多数应用都是两层模型且层2样本量较大。HLM方法在广泛应用的同时仍存在忽略前提假设检验、分析过程中的重要信息和结果报告不完整等问题,随后提供了4条建议。

To explore the present situation and problems of using the hierarchical linear modeling （HLM） in psychological research in China, this article reviewed 50 studies published in Chinese psychological journals, retrieved from the CNKI database from 2002 to 2011. A checklist was derived in terms of the methodological literature on hierarchical linear modeling for selecting the candidate stud- ies, which focused on the issues of model development and specification, data preparation, estimation method and hypothesis testing. The resuhs of survey shoued that, first, the HLM has been widely used in the broad psychological sub - fields in Ctfina, such as manag- erial, developmental and educational psychology. Second, the most common applications werethe two - level models where individuals were nested within contexts. It is worth pointing out that the sample size of level 2 is significantly larger than the sample size of level 1. Although theHLM in the psychological fields has been widely used, there are a number of potential problems in its application. First, in model development and specification, one could not determine how many models were estimated in five articles, what type of centering if any was used in most of the articles, whether the null model was tested in 50% of the articles. Second, in data preparation, no article reported whether the data were consistent with assumptions, for example, distributional assumptions; no article implements the prior power analysis. In many articles, one could not determine whether the data were complete ; and if they were not complete, the methodology responsible for the incomplete datawas not described. Third, in the estimation of parameters and hypothesis testing, one could not, except for six studies, determine how the models were estimated. Inappropriate and insufficient information were reported in most of the studies. For example, thestandard errors for the parameters of interest and proportion of explained variance were missing. Finally, some guidelines for reporting hierarchical linear analyses are provided at the end of the paper.