Prof. J. K. Vermunt, Dr J. Mulder & Dr D. L. Oberski
Advancing structural equation modeling with unbiased Bayesian methods
Bayesian structural equation modeling (SEM) is becoming increasingly popular in applied research as an alternative to classical SEM. In the Bayesian approach, a prior needs to be specified. When appropriately chosen, the prior yields higher statistical power, prevents technical problems occurring in classical SEM, such as nonconvergence and inadmissible solutions, and allows the researcher to incorporate state-of-the-art substantive knowledge. When inappropriately chosen, however, priors cause bias. The goal of this project is to develop novel priors for Bayesian SEM that overcome the technical limitations of classical SEM while avoiding bias. The new methodology will be implemented in user-friendly statistical software.
NWO Research Talent Grant
1 September 2015 – 1 September 2019