A Comparison of the Diagonal and Cross-Sectional Designs When Assessing Longitudinal Mediation

Mediational studies are of interest in psychology since they explain the underlying relationship between two constructs. The gold standard for studying mediation in psychology is a longitudinal latent variable model. However, many mediational studies use a cross-sectional approach even though deficiencies have been identified. In recent publications, a new approach, the diagonal (sequential) design has been used to examine mediation. This design is a compromise between the cross-sectional and longitudinal designs since it incorporates time in the model but only has one measurement of X, M, and Y. These two designs were compared to see whether the diagonal design assesses longitudinal mediation more accurately than the cross-sectional design when mediation is full or partial. It was found that the diagonal design does not assess longitudinal mediation more accurately than the cross-sectional design. Neither designs cannot be depended on to accurately assess longitudinal mediation.