The Effective Statistician - in association with PSI Structural Equation Models
Feb 5, 2024
Christian Geiser, expert in structural equation modeling, discusses the theoretical foundations and diverse applications of SEMs. He explores latent variables, factor analysis, and different software options for working with SEMs. He also shares insights on his career move and founding a statistics training company. Finally, he provides valuable resources such as workshops, courses, and books for learning SEMs.
AI Snips
Chapters
Transcript
Episode notes
Structural Equation Models Overview
- Structural equation models (SEMs) generalize regression by analyzing multiple dependent and independent variables simultaneously.
- SEMs can model latent variables to separate true score variance from measurement error, improving precision.
Measurement Error Bias in Estimates
- Measurement error in exposure or outcome variables leads to biased regression coefficients, often underestimating effects.
- SEMs help correct this bias by accounting for measurement error, improving estimate accuracy, especially in fields like epidemiology.
Concrete Meaning of Latent Variables
- Latent variables represent the true score underlying observed variables, defined psychometrically as the expectation over repeated measurements.
- These variables allow separating reliable variance from error in scales like those measuring anxiety.