How scientists measure what they cannot see
In short: How do you quantify something as invisible as social class or intelligence? Researchers use Structural Equation Modelling (SEM) to bridge the gap between observable data and hidden, latent constructs, allowing us to estimate the true, error-free reality behind imperfect measurements.
Structural Equation Modelling (SEM) is a multivariate statistical framework used to model complex relationships between directly observed variables and indirectly observed, or 'latent', variables. These latent constructs—such as social capital, trust, or even mental illness—are intellectual devices used to categorize phenomena that are not directly observable. By treating a latent variable as a symbol for a true, error-free variable, SEM allows researchers to quantify the measurement error present in the indicators used to estimate it.
The power of SEM lies in its integration. Unlike standard regression, which may only evaluate partial or constrained models, SEM can represent complex theories in a single, integrated model. It functions as a combination of linear regression and factor analysis, performing all measurements and tests simultaneously in one estimation procedure. This simultaneous approach can lead to more accurate estimates than if each component were calculated separately.
The methodology is increasingly vital for validating administrative data. For example, the Office for National Statistics (ONS) recently used SEM to investigate measurement error in floor area data from the Valuation Office Agency (VOA) and Energy Performance Certificate (EPC). Their analysis revealed that VOA data tended to have less measurement error than EPC data at a national level. While SEM is a robust tool for assessing data quality without needing a 'gold standard' measurement, researchers must remain cautious, as results still reflect the specified model and may contain potential bias.
Opens the Intellect+ feed with this card first. Swipe up for the next one.
Sources
- Summary by Intellect · source: Using structural equation modelling to determine measurement error in different administrative data sources - Office for National Statistics
- Also: Structural equation models | methods@manchester | The University of Manchester
- Also: Structural Equation Modeling - PubMed
- Also: Structural equation modeling
To cite this card, link to https://intellectplus.co.uk/app/health-structural-equation-modeling.
How we source, check and correct cards: Editorial standards.
General information, not professional advice. Spotted an error? Flag the card or see corrections and complaints.
Intellect Plus is a visual educational discovery platform for exploring fascinating facts and ideas. Brain Food, Not Brain Rot. What is Intellect Plus?
Related
More in Statistics · More in Ideas & Philosophy · Latest · A–Z · All topics
