Structural Equation Modeling found to be effective for continuous time models with small data sets

When researchers need to organize data for a study, they often turn to different methods of modeling to analyze and predict outcomes from their data. One such method is Multilevel Latent Differential Structural Equation Modeling, also known as M-LDSEM, which is a method for estimating continuous time models within a traditional structural equation modeling (SEM) framework. 

According to new research from a team at Penn State, M-LDSEM is effective at analyzing and understanding longitudinal data even when there are a small number of measurements, such as 10 or 20 days. 

Continuous time models are useful for studying changes over time with longitudinal data, which is gathered from a sample at different points in time. These models also help researchers handle data with missing values or varying time intervals, which are common in intensive longitudinal studies. 

By using a multilevel framework, M-LDSEM can identify outcomes within a single individual and between multiple individuals in data collected from a participant group over time. Ultimately, the M-LDSEM method can help researchers understand how two processes influence each other over time, when applied to coupled dynamic models. 

“We expect this model to impact fields such as psychology and behavioral research,” said Young Won Cho, a graduate student in human development and family studies at Penn State, and lead author on the paper. “Unlike physics or economics, behavioral research often requires significant effort and resources to gather data. Although not without limitations, the fact that this model can function effectively with as few as 10 days of data is encouraging. This could enable more scientists in the behavioral research field to utilize continuous time models. 

“It is rewarding that this work contributes to bridging the gap between empirical research and advanced modeling methods,” said Cho, who is supported by her adviser, Sy-Miin Chow, professor of human development and family studies, another author on the paper. 

Lynn Martire, professor of human development and family studies and senior author on the paper, said the findings have a direct impact on the research performed in her lab, The Couples and Health Lab. 

“Our lab is focused on the impact of chronic illness symptoms and their long-term emotional consequences on the spouse, as well as how spouses’ emotions and behaviors affect the experience of chronic illness,” Martire said. “The findings from this paper suggest that an M-LDSEM approach can help identify which individual has the stronger influence on their partner under specific conditions, ultimately helping us improve health and wellness for those living with chronic illnesses.”

Originally published in August 2024.