Developmental Methodology
Studying the interplay between peer relationships and problem behavior requires applying and refining the best statistical tools. In particular, recent advances in latent variable modeling (e.g., Item Response Theory, Latent Class Analysis), social network analysis, multilevel modeling, and Bayesian statistics have potential applications in developmental science. Unfortunately, innovations in statistics are often slow to cross disciplinary boundaries. Improving the transfer of new methods from statistics to the social sciences is critical, because applying the wrong models can result in misleading or wrong conclusions. For example, early research concluded that peers strongly influenced each other, but this work often ignored selection effects and network structure or assumed influence from cross-sectional correlations.
The reciprocal is also true: It is critical that methodological models be informed by developmental and sociological research. For example, social network analysis can identify how people are connected and suggest how diffusion processes may work, but this research often ignores the meaning of the ties and the individual characteristics of the “nodes” (or people) involved, both of which might affect how influence spreads. In addition, IRT models that are appropriate for certain educational applications may need to be adapted when applied to study psychopathology or delinquency.
Click on any of the pictures below to read more about my specific research interests in developmental methodology: