Multilevel Covariance Structure Analysis of Sibling Antisocial Behavior

Conventional covariance structure analysis, such as factor analysis, is often applied to data that are obtained in a hierarchical fashion, such as siblings observed within families. Multivariate modeling of such data, however, is most frequently done as if the data were obtained as a simple random sample from a single population. An alternative specification is presented that explicitly models the within-level and between-level covariance matrices in familial antisocial behavior. Sibling data from the National Youth Survey, a national probability sample of youth, were used to specify a multilevel covariance structure analysis of sibling antisocial behavior. Results demonstrate homogeneity in antisocial behavior within sibling clusters but heterogeneity across families. These analyses highlight potential pitfalls of ignoring issues of independence and demonstrate how conventional covariance structure software can be easily adapted to handle hierarchical models, providing a large set of new analysis possibilities for multilevel data.