Causal inference is at the heart of many questions in social sciences, yet the study of it is in some sense still a novel and currently thriving field. In our lab we draw on the classic Campbellian tradition, the potential outcomes framework, and causal graphical models. In particular we work on extending methods of propensity score matching to complex data situations and improve model testing in causal graphical models.
Missing data are a prevalent problem in nearly all studies conducted in psychology, education, and other fields. Principled methods to deal with missing data are slowly being adapted by these fields. However, there are still a host of unresolved issues concerning selection of so-called auxiliary variables, or methods to deal with missing data in complex data analytic situations, e.g. those involving clustered data or latent heterogeneity.
Structural equation models
Structural equation modeling is a flexible statistical tool to model (among other things) latent factors and direct and indirect effects. Recently there has been more interest in the causal underpinning of such models, in particular the causal assumptions involving indirect effects. Our lab researches methods that strengthen causal conclusions of such models in the presence of causal heterogeneity. Furthermore, we are interested in exploring rigorous tests of models that allow eliminating alternative causal pathways.
Mabry, L., Elliot, D.L., Kuehl, K.S., Kuehl, H., Moe, E.L., Goldberg, L., DeFrancesco, C.A., MacKinnon, D.P., Favorite, K.C., & Thoemmes, F.(2014). Adult health education in the workplace: The adoption, benefits, and durability of PHLAME. Proceedings of the Ireland International Conference on Education, Dublin, IR.