Statistical procedures commonly applied to behavioral data, such as factor analysis, structural equation models, and mixed-effects models, fall under a general statistical framework based on the means, variances and covariances of data. My research concerns extensions and applications of this general framework to the study of clustered, repeated measures, and longitudinal data.
For data observed over time, researchers are often interested in characterizing normative change, as well as studying individual differences in change by documenting evidence that individuals differ in their degree or form of change and possibly attempting to account for individual differences by including person-level or contextual variables in the model. Popular methods for the study of clustered data, commonly known as hierarchical linear models, multilevel models, and mixed-effect models, are routinely applied to repeated measures and longitudinal data. A general difference in the application is that models for repeated measures and longitudinal data typically involve functions of time, in addition to other important predictors. A desirable feature of this framework is that individuals need not be observed at the same time points and missing data are, in general, easily handled. In addition, many different forms of change may be modeled.
Current research centers on mixed-effects models when data are not missing at random. Appropriate application of methods for missing data is essential in behavioral research. My work concerns the application and evaluation of methods to address mechanisms that give rise to missing data in multivariate analysis.
In addition to methodological work, I have developed a program of research based on health and social psychology. This has generated a line of research in collaboration with colleagues working in areas such as adolescent health, well-being, chronic illness, and drug and alcohol use, as well as health education and preventative medicine, health psychology, animal behavior, and nursing.