Structural equation modeling (SEM) combines the elements of factor analysis and path analysis to evaluate both measurement and structural models simultaneously. The measurement model correlates the observed indicators (manifest variables) to their latent constructs while the structural model examines the relationships between those formed latent variables. While SEM is commonly applied in social and behavioral sciences, it is not as frequently used in other research fields for modeling.
Recently, Katie Sternberger, MS, a new graduate of our Biostatistics & Data Science program, demonstrated the use of SEM models at the Conference on Statistical Practice (CSP) 2024, using data from the Women’s and Their Children’s Health study. She concluded that SEM is a powerful and flexible statistical analysis technique, particularly useful when dealing with latent variables. Her co-authors included her advisor, Dr. Evrim Oral, along with Drs. Ariane Rung, Nicole Nugent, Edward Peters and Edward Trapido.