Methodology and Statistics
Tilburg School of Social and Behavioral Sciences
Structural equation modeling as an antidote to selective outcome reporting
A prevalent concern in psychology is the issue of selective reporting of dependent variables or outcome measures. It is expected that many researchers, when confronted with diverging significance results, to be tempted to focus only on those dependent variables that show significance, leading to selective outcome reporting. Selective outcome reporting based on significance obscures effects that are smaller yet potentially relevant for theory or practice, and inflates effect sizes in meta-analyses of combined effects. As an antidote to selective outcome reporting, a Multi-Group Confirmatory Factor Analysis (MGCFA) model with mean structure and measurement invariance constraints can be fitted on means and covariances within conditions, which makes focusing only on those dependent variables that show significance unnecessary.
In this project we use MGCFA with mean structure as a method for analyzing multivariate experimental data, develop it for various common experimental designs and mediation analyses, study small sample behavior and power to detect violations of invariance and latent effects, disseminate the method by developing useful software, and illustrate the method by submitting it to real data.
Prof. dr. J.M. Wicherts, Prof. dr. M.A.L.M. van Assen
ERC Consolidator Grant
1 September 2017 – 31 August 2021