Quantitative lunch -- Seohyun Kim (Zoom)

o Finite mixture modeling is a statistical method that describes an unknown distribution using a mixture of known distributions and has shown to be useful in modeling heterogeneous data with a finite number of subpopulation. Finite mixture modeling has been used in various contexts including psychological research in identifying qualitatively distinct groups of individuals in the population of study. In this talk, I will introduce two studies that use a finite mixture modeling framework. The first study introduces a robust approach for growth mixture modeling that is less sensitive to outlying observations. Growth mixture modeling is a combination of growth curve modeling and finite mixture modeling, and it is a popular analytic tool for longitudinal data analysis. In this study, I use a median regression approach to ensure that the parameters of growth mixture modeling are less influenced by extreme observations. The second study introduces a model that addresses text data paired with numbers, such as essays and essay scores. This model is designed to detect meaningful subgroups that are determined by the relationships between examinees’ responses and their scores. This model is illustrated by a real data analysis from a National Science Foundation-funded study on teaching science to both English-language learners and native English speakers.