Including auxiliary cases to address missing data issues through multiple-group models

This study explores the use of auxiliary cases to address missing data problems. Extending from the framework of integrative data analysis (IDA), auxiliary cases are defined as cases from other independent samples or existing studies that, while not part of the target group of interest, have sufficient data on the target variables to reduce the impact of missing data on target analyses. These cases can be included in analyses through multiple-group models as an auxiliary group with proper cross-group measurement invariance constraints, while allowing focal structural parameters to be freely estimated. Although past studies based on complete data have found that IDA can stabilize the estimates, most IDA studies mainly consider missing data as a challenge of data integration. In this study, we offer a different perspective by showing that data integration can also serve as a method to limit the impact of missing data. Specifically, our simulations show that the effects of including auxiliary cases on the power and efficiency of the target analyses are more substantial under missing data conditions than under complete data conditions. However, if there are improper cross-group invariance constraints in the analytic model or if inter-dataset heterogeneities are not accounted for in the analyses, the inclusion of auxiliary cases may instead introduce additional bias and reduce power, especially when full information maximum likelihood is used. An empirical example based on two independent experimental datasets is provided. Some issues related to the use of auxiliary cases are also discussed.