Nidhi Kohli, University of Minnesota

Nonlinear random-effects models for observed, continuous longitudinal data are often used in education, psychology, and broader social sciences to examine individual- and population-level curvilinear development (or growth) over time. The piecewise function is a popular and flexible intrinsically nonlinear function for analyzing segmented trends in individual trajectories over time. Piecewise random-effects models (REMs) allow each segment of the overall developmental trajectory to have a different functional form (e.g., linear-linear, quadratic-linear). The random knot (changepoint), one of the most interesting parameters of the model, is the unknown timepoint of transition from one developmental segment to another. The statistical framework of piecewise REMs has been extended in several ways to enable researchers to analyze different research questions. In the first extension, the piecewise REM allows the detection of multiple knots where data come from a mixture of two or more sub-populations (latent classes). Furthermore, covariates can be incorporated to aid in the identification of latent classes. In the second extension, the model allows for the bivariate modeling of piecewise trajectories for two interdependent longitudinal outcomes (e.g., modeling mathematics and reading achievement scores). In the third and last extension, the model can statistically capture the impact of both dynamic and static group membership on individual outcomes. Each methodological extension is motivated by empirical data applications.

about the speaker

Dr. Nidhi Kohli is a Yackel Professor of Educational Measurement and Assessment in the Quantitative Methods in Education program of the Department of Educational Psychology in the College of Education and Human Development at the University of Minnesota. The overarching focus of her methodological research interests includes, but are not limited to, the development and application of novel statistical methods for analyzing a variety of longitudinal (i.e., repeated measures) data in the areas of education, psychology, and, more generally, social and behavioral sciences. She is the Editor of the Application Reviews and Case Studies section of Psychometrika, and currently serves on editorial boards of Journal of Educational and Behavioral Statistics, Multivariate Behavioral Research, Psychological Methods, Educational and Psychological Measurement, and Journal of Educational Psychology.