Review of the Application of Hierarchical Linear Modeling in Psychological Research from 2002 to 2011 in China
方杰,邱皓政,张敏强,方路
DOI: https://doi.org/10.16719/j.cnki.1671-6981.2013.05.033
2013-01-01
Abstract: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 studies,which focused on the issues of model development and specification,data preparation,estimation method and hypothesis testing.The results of survey shoued that,first,the HLM has been widely used in the broad psychological sub-fields in China,such as managerial,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.