Strategies for Multidomain Sequence Analysis in Social Research

Abstract

Multidomain/multichannel sequence analysis has become widely used in social science research to uncover the underlying relationships between two or more observed trajectories in parallel. For example, life-course researchers use multidomain sequence analysis to study the parallel unfolding of multiple life-course domains. In this article, the authors conduct a critical review of the approaches most used in multidomain sequence analysis. The parallel unfolding of trajectories in multiple domains is typically analyzed by building a joint multidomain typology and by examining how domain-specific sequence patterns combine with one another within the multidomain groups. The authors identify four strategies to construct the joint multidomain typology: proceeding independently of domain costs and distances between domain sequences, deriving multidomain costs from domain costs, deriving distances between multidomain sequences from within-domain distances, and combining typologies constructed for each domain. The second and third strategies are prevalent in the literature and typically proceed additively. The authors show that these additive procedures assume between-domain independence, and they make explicit the constraints these procedures impose on between-multidomain costs and distances. Regarding the fourth strategy, the authors propose a merging algorithm to avoid scarce combined types. As regards the first strategy, the authors demonstrate, with a real example based on data from the Swiss Household Panel, that using edit distances with data-driven costs at the multidomain level (i.e., independent of domain costs) remains easily manageable with more than 200 different multidomain combined states. In addition, the authors introduce strategies to enhance visualization by types and domains.