Pub. date: 2009 | Online Pub. Date: October 05, 2009 | DOI: 10.4135/9780857020994 | Print ISBN: 9781412930918 | Online ISBN: 9780857020994| Publisher:SAGE Publications LtdAbout this handbook
Chapter 24: Multilevel Latent Variable Modeling: Current Research and Recent Developments
David Kaplan & Jee-Seon Kim & Su-Young Kim
Multilevel latent variable modeling: Current research and recent developments There are many instances of research problems in the social and behavioral sciences where observations are not simple random samples from some defined population. For example, organizations such as schools are hierarchically structured, and the data generated from these types of organizations are typically obtained through some form of multistage sampling. Ignoring the sampling structure through the disaggregation or aggregation of data derived from such structures is fraught with problems. The difficulty with disaggregation or aggregation is they are not optimal approaches for a proper analysis of the actual structure of the data. Using students in schools as an example, the problem with disaggregation is that students will have the same values on observed and unobserved school level variables. As such, the usual regression assumption of independence of errors is violated, possibly leading to biased regression coefficients. In the case of ...