Pardo Antonio, Ruiz Miguel A, San Martín Rafael
Universidad Autónoma de Madrid.
Psicothema. 2007 May;19(2):308-21.
Hierarchic or multilevel models are used to analyse data when cases belong to known groups and sample units are selected both from the individual level and from the group level. In this work, the multilevel models most commonly discussed in the statistic literature are described, explaining how to fit these models using the SPSS program (any version as of the 11 th ) and how to interpret the outcomes of the analysis. Five particular models are described, fitted, and interpreted: (1) one-way analysis of variance with random effects, (2) regression analysis with means-as-outcomes, (3) one-way analysis of covariance with random effects, (4) regression analysis with random coefficients, and (5) regression analysis with means- and slopes-as-outcomes. All models are explained, trying to make them understandable to researchers in health and behaviour sciences.
当案例属于已知组且样本单位是从个体层面和组层面进行选择时，层次模型或多级模型用于分析数据。在这项工作中，描述了统计文献中最常讨论的多级模型，解释了如何使用SPSS程序（第11版及以后的任何版本）拟合这些模型以及如何解释分析结果。描述、拟合并解释了五个特定模型：（1）具有随机效应的单向方差分析，（2）以均值为结果的回归分析，（3）具有随机效应的单向协方差分析，（4）具有随机系数的回归分析，以及（5）以均值和斜率为结果的回归分析。对所有模型都进行了解释，力求让健康与行为科学领域的研究人员能够理解。
