When the variables of interest are measured by a set of items on units having a multilevel setting, conventional structural equation models cannot be used because the assumption of independence of all latent variables and indicators across units is violated due to the within-cluster dependence. In this work we propose the use of parcelling in defining of latent variables of a multilevel structural equation model (MSEM). The paper aims to face the problem of the use of categorical item response data when a multilevel SEM must be applied.
|Title of host publication||Statistical Methods for Service Quality Evaluation - Book of short papers of IES 2019, Rome, Italy, July 4-5|
|Number of pages||5|
|Publication status||Published - 2019|