My research

I am interested in structural equation modeling (SEM) in general, and specifically in combination with measurement invariance, multilevel data, or meta-analysis (MASEM).

My dedication to research on MASEM started when I was working on my PhD on multilevel SEM. My attraction to MASEM was due to both the broad range of research fields for which MASEM can be extremely helpful, as well as the intrinsic beauty of the MASEM method. MASEM combines using the most informative data with a technique that is so versatile that many other methods can come to be viewed as a form of SEM.

Ever since I was intrigued by MASEM, the goal of my research has been to make the technique better known and more user friendly for researchers, and to develop new MASEM methods where needed. I published a book on MASEM (Book) that serves as an introduction, overview of methods, and practical guide for the specific MASEM method called Two-Stage SEM.

After obtaining my PhD in multilevel SEM, I was awarded an NWO Rubicon grant to expand my scientific horizon and further develop MASEM at the National University of Singapore with prof. dr. Mike Cheung. In the Rubicon-project, I developed a method to handle missing correlation coefficients in MASEM, and a method to test differences in MASEM-parameters across subgroups of studies.

From 2016, an NWO Veni grant from the Dutch Research Council allowed me (in collaboration with Mike Cheung) to develop a new MASEM method; one-stage MASEM. The new method formed a breakthrough in the explanation of heterogeneity across studies in MASEM. The methods so far could only compare SEM parameters across subgroups of studies. As a result, continuous variables had to be categorized, leading to a loss of information and power. The one-stage MASEM method that I developed allows for the analysis of continuous study-level moderator variables in MASEM. Moreover, one-stage MASEM has attractive statistical properties because it is a one-stage method that uses maximum likelihood estimation throughout.

Another line of my statistical research focuses on multilevel SEM. Multilevel SEM is the appropriate technique if one wants to fit SEMs to clustered data, such as data from children nested in schools. Here too, one of my main aims is to bring multilevel SEM within reach of researchers in various fields, by writing instructional articles, and by providing stepwise procedures and example syntaxes. I produced models and instructions for testing measurement invariance in multilevel data, and explanations of the relationship between measurement invariance across clusters, invariance across levels, and reliability in multilevel SEM.