Andres Felipe Perez Alonso

Methodology and Statistics
Tilburg School of Social and Behavioral Sciences
Tilburg University

Mixture multigroup Structural Equation Modeling for comparing latent variable relations among many groups

Social scientists often examine the relationships or mechanisms between two or more latent variables, and Structural Equation Modeling (SEM) is the state-of-the-art for doing so. In case of many, researchers may also be interested in finding subsets of those groups that share similar relationships. In other words, a research problem can involve understanding a latent relationship between unobserved variables while, at the same time, searching for similarities and differences across a large number of groups based on that same mechanism (Brandt et al., 2021).

For validly comparing the latent variables’ relations among groups, the measurement of the latent variables should be invariant across the groups (i.e., measurement invariance). However, often, at least some measurement parameters differ when comparing many groups. Restricting these measurement parameters to be equal across groups, when they are not, causes the structural relations to be estimated incorrectly and thus invalidates their comparison (Guenole & Brown, 2014; Pokropek et al., 2019). Current mixture clustering SEM methods restrict those measurement parameters within clusters when searching for subsets of groups (Kim, Cao, Wang, & Nguyen, 2017). Thus, they do not allow for clustering groups on their structural relations specifically, but rather capture measurement non-invariances with the same clustering.

To capture differences and similarities in structural relations while accounting for the reality of measurement non-invariance, this project aims to develop a mixture multigroup SEM (MixMG-SEM) framework. MixMG-SEM obtains a clustering of groups focused entirely on the structural relations by making them cluster-specific, while allowing for the measurement parameters to be (partially) group-specific. In this way, MixMG-SEM disentangles differences in structural relations from differences in measurement parameters.

Dr. Kim de Roover
Prof. Dr. Jeroen Vermunt
Prof. Dr. Yves Rosseel

Financed by
NWO Vidi (granted to Kim de Roover)

1 December 2021 – 30 November 2025