I have three principal research interests: (1) multivariate methods for the analysis of change, (2) using multiple group and latent class models to understand divergent developmental processes, and (3) the development and application of machine learning methods for psychological science.
My research in this area focuses on methods to analyze repeated measures data to evaluate long-term systematic trends and between-person differences therein. Such data are typical in the study of developmental changes, such as changes in mathematics, reading, behavior problems, and depression. These sorts of data often show systematic patterns of change; however the pattern and amount of change often vary over people making modeling of these types of data more complex. My research in this area has focused on model specification (Grimm, 2007; Grimm & Liu, 2016; Grimm & Marcoulides, 2016; Grimm, Ram, & Hamagami, 2011; Grimm & Widaman, 2010; Ram & Grimm, 2007), nonlinear forms of change (Grimm & Ram, 2009; Grimm, Ram, & Estabrook, 2010; Grimm, Ram, & Hamagami, 2011; Grimm, Zhang, Hamagami, & Mazzocco, 2013), and latent change score models (Grimm, 2012; Grimm, An, McArdle, Zonderman, & Resnick, 2012; Grimm, Castro-Schilo, & Davoudzadeh, 2013; Grimm, Zhang, Hamagami, & Mazzocco, 2013; McArdle & Grimm, 2010).
My research in this area focuses on models for examining heterogeneity in development. The growth models allows for a specific type of heterogeneity as the variability in latent intercepts and slopes is normally distributed. Growth mixture models, a combination of the finite mixture model and growth model, allow for heterogeneity to be examined in terms of latent classes with divergent developmental trajectories. My work in this area has focused on model specification (Grimm, McArdle, & Hamagami, 2007; Ram & Grimm, 2009), the incorporation of measurement models to aid in the determination of latent classes (Grimm & Ram, 2009), modeling nonlinear trajectories with multiple latent classes (Grimm, Ram, & Estabrook, 2010; Serang, Zhang, Helm, Steele, & Grimm, 2015), and model selection (Grimm, Mazza, & Davoudzadeh, 2017; Ram & Grimm, 2009; Grimm, Houpt, & Rodgers, 2021; Houpt, Grimm, McLaughlin, & Van Tongeren, 2024).
Machine learning methods are not necessarily well suited for psychological science where our statistical models involve unmeasured (latent) variables, our theories involve indirect effects, and our data have dependency due to repeated measurement or clustering. My research in this area has focused on the combination of data mining methods with statistical models used in psychological science. This work can be seen in Jacobucci, Grimm, and McArdle (2016) where regularized regression was combined with structural equation models, Serang, Jacobucci, Brimhall, and Grimm where lasso regression was incorporated into mediation models, and Grimm, Mazza, and Davoudzadeh where k-fold cross-validation was used for model selection in mixture models. We have worked on recursive partitioning approaches for nonlinear mixed-effects models (Stegmann, Jacobucci, Serang, & Grimm, 2018), the development of more efficient recursive partitioning algorithms for use with latent variable models (Serang, Jacobucci, Stegmann, Brandmaier, Culianos, & Grimm, 2021), missing data algorithms for data mining methods (Rodgers, Jacobucci, & Grimm, 2021), and the development of new recursive partitioning algorithms for psychological data (Grimm & Jacobucci, 2021).