Extraction of latent class progressions in a longitudinal study with applications in educational and behavioral science data

Jennifer Lord-Bessen, Fordham University


In the study of developmental trajectories, classifying growth composed of stages best represented by categorical variables is not currently feasible. Growth Mixture Modeling can only be applied to data where an underlying functional form is assumed. Latent Transition Analysis can identify the likelihood of transitioning between stages but only between two time points. ^ This study developed and tested a new exploratory latent class methodology for repeated measures data, Latent Class Progression Analysis (LCPA), which identifies different subgroups' developmental trajectories over three or more time points where the trajectories are in terms of a progression through levels of a discrete latent variable. LCPA provides a way to classify and describe growth when the data is not continuous in nature. ^ A series of simulation studies was conducted to explore the performance of LCPA using scenarios drawn from both education and behavioral science. The method identified most of the typical developmental trajectories (progressions) in the sample accurately. In addition, it recaptured subjects' simulated stages at each time point as well as their likelihood of answering each item correctly conditioned on their stage. Two extensions to the method were investigated. An extension that changed the method of latent class assignment from modal to proportional did not result in improvement in either classification accuracy or extraction of progressions. A second extension to the method using covariates was shown to improve classification accuracy and increase the number of progressions identified within the sample. An application to the behavioral dataset investigating adolescents' change in sexual habits and risky behaviors demonstrated that LCPA could extract meaningful progressions from a substantive area. ^ This methodology has multiple applications in education and behavioral science. Within learning progression research, how students learn a subject entails identifying the stages of learning and the order different groups of students pass through them. In the study of behavioral change, researchers are interested in questions involving change in moods and other psychological states. With the development of LCPA, researchers may now address these and similar questions where change in categorical variables across three or more time points is of interest.^

Subject Area

Quantitative psychology

Recommended Citation

Lord-Bessen, Jennifer, "Extraction of latent class progressions in a longitudinal study with applications in educational and behavioral science data" (2014). ETD Collection for Fordham University. AAI3674002.