Missing Data in Practice with Structural Equation Models

When using survey data, researchers must evaluate how to effectively handle missing data. For social survey data, full information maximum likelihood methods are often implemented when the researcher is interested in structural equation models. This strategy is convenient to implement and provides acceptable results, yet it does not incorporate any imputation methods for assessing the missing information. We examine the benefits of imputation methods as an alternative for managing missing data, particularly in longitudinal surveys where missingness may be conditioned on previous panel data. In this talk, I will outline methods for handling missing data and briefly discuss structural equation models. We then apply these procedures to the Family Transitions Project, a longitudinal survey of more than 550 participants, which focuses on familial relationships and socioeconomic stress induced by economic hardships.

Event Topic:

Computational Mathematics & Statistics