There is an ongoing debate about the best way to approach measurement of psychological constructs in structural equation modelling (SEM).

The default psychometric approach is assuming a reflective measurement model that can account for measurement error, but recent work suggests that instead using a formative measurement approach might have both epistemic and pragmatic advantages.

In this project, we investigate the implications of using different measurement models in a misspecified model where the underlying structural model is a causal graph instead of a latent variable.

We systematically vary the network structure and parameters to investigate how the choice of measurement model influences the inferences drawn from the data.

The aim is to provide a more nuanced view on the application of different measurement models in SEM mediation models, and provide some insights into the use of formative and reflective measurement models for summarising complex psychological constructs.