ABSTRACT
In recent years a number of statistical techniques and programs for the analysis of longitudinal multilevel data have become available, allowing researchers the opportunity to more adequately address questions related to development and change and to simultaneously study behavior at different levels (e.g., individual, family, school, and neighborhood levels). A l - though many of these statistical approaches are widely available, they are still relatively underused, and there is uncertainty among researchers as to the appropriateness and usefulness of different approaches for analyzing longitudinal data. The purpose of this chapter is to compare three analytic approaches to answering longitudinal multilevel research questions using unbalanced adolescent and family alcohol use data. The analytic approaches include (a) a full information maximum likelihood (FIML) latent growth modeling (LGM) approach using an extension of a factor-of-curves model, (b) a limited information Multilevel L G M ( M L G M ) approach using Muthén 's ML-based estimator ( M U M L ) , and (c) a full information hierarchical linear modeling approach (HLM) . Data are from the National Youth Survey (NYS) and comprised 888 adolescents (443 males and 445 females; mean age = 13.86 years) from 369 households. Results demonstrate similarity in outcomes and interpretations derived from the three analytic techniques. Discussion includes comparison of the three techniques, including advantages and limitations.
