| Mediation analysis is a statistical method to explore the relationship among variables,which provides a statistical technique for revealing the complex relationships among variables,and has played an important role in many fields in recent years.Longitudinal data analysis is one of the leading issues of statistical research recently,clinical medicine,often requires long-term tracking of individual data.Because of the repeated measurement is nested in the same individual,it can also be regarded as the multi-layer structure data.Multilevel linear model can explain the nested structure data,so the multilevel mediation model,which combined the multilevel linear model and the mediation analysis model,can be a good way to reveal the internal impact mechanism among longitudinal data.The research on the mediation analysis is scattered presently,and the methods are outdated.So it is necessary to summarize the existing literature.Besides,the existing research mostly based on the cross-sectional data,but less researches on the longitudinal data.Especially the sample size determination,which is crucial for the mediation studies,If the sample size is too small,the study may lack adequate statistical power,if sample size selected is too large,it is not only a waste of the resources but also slow down the research process.Based on the above problems,this paper systematically combed and summarized the basic theory about simple mediation model and multilevel mediation model firstly,meanwhile recomends the latest process of mediation analysis.Moreover,this paper also introduces the mediation with binoary dependent variable.Secondly,this article provides the sample size required to achieve 80% power when using Sobel's and distribution of product methods for longitudinal mediation studies under various sizes of mediation effect,within-subject correlations and the number of repeated measures.The conclusion can provid guidance in choosing sample sizes for mediation studies with longitudinal data. |
PDF Full Text Request