Multilevel methods for modeling observed sequences of family interaction

Abstract

Observation of interaction plays a central role in family research. This article discusses how to analyze sequential data generated by discrete microcoding methods to test hypotheses about family interaction. Current methods for studying sequential data are presented, and their limits are discussed. Building on recent applications of contingency table analysis to such data, a multilevel log-linear model is presented that can specify and estimate indicators of individual behavioral tendencies and antecedent-consequent relationships among behaviors, both within and across samples of families. An example of this method is presented using data from a study of couples facing job loss. Potential extensions of this framework for future research are discussed.