Data in physical educational research often have a multilevel structure, with observations on individuals nested within groups. Such data have been commonly analyzed at the individual level, whereas the group level variation has been ignored. The purpose of this article is to illustrate the use of multilevel confirmatory factor analysis via a didactic example of a two-level (students nested within classes) analysis of six items test battery artificial data. The results of this analysis indicate that a two-factor model adequately described within-class (individuals) differences in tests results. By contrast, at the between-class level (groups), a general factor of battery was sufficient to capture class differences in test results. In addition, two approaches to estimating population between- and within-class covariance matrices are discussed here. Present article demonstrated the merit of using multilevel techniques in researches, where data have multilevel components. Finally, several software programs appropriate for multilevel confirmatory factor analysis are recommended here.

Charles University Faculty of Physical Education and Sport Prague

Víceúrovňové rozšíření faktorové analýzy: důležité aspekty pro tělovýchovný výzkum

Lit.: 27