Considering measurement models can enhance the analysis of multivariate dyadic data

What is it about?

Dyadic data contain rich information about dynamic relationships that occur within a pair of people. The Actor-Partner Interdependence model (APIM; Kenny, 1996) has been widely used to study the interdependence of dyad members, defining its patterns by the relative strength of influence that members receive from their partners. This study proposes to add measurement models to the APIM to analyze multiple correlated variables as manifestations of underlying theoretical constructs that are interrelated between dyad members. In addition to this latent APIM approach, two other extensions of the APIM for multivariate data are presented in comparison: Instead of including measurement models, the manifest APIM uses observed variables simultaneously, and the composite-score APIM analyzes sums of observed variables. The usage of the three methods and their interpretations are presented with publicly available data from married couples, and practical recommendations for applied researchers are provided based on the findings from a simulation study.

Featured Image

Why is it important?

While the choice of analysis models may seem technical, our findings show that each of the three methods may lead to substantially different conclusions depending on the context in which the variables in the analysis are related to each other. Thus, we provide a list of recommendations on how to choose and apply more suitable APIM approaches in practice and discuss the importance of consulting substantive theory and measurement characteristics when selecting a method to analyze multivariate dyadic data.