I am learning and have used Structural Equation Models with latent variables, with measurement models and path analysis.

It seems however, that I should be able to use them even without latent variables-- maybe if I just conduct the path analysis without looking into measurement models, which would not be required in the absence of latent constructs.

I would also be grateful for any reading material on this, both theoretical and applied.


1 Answer 1


Yes, you can do structural equation models without latent variables.

Regression, t-tests (paired and unpaired) can all be considered to be SEMs without latent variables. In addition, things like mediation analysis, or cross-lagged regression analysis, can also be done as SEMs, without (or with) latent variables.

(Well, in a sense that's not true. Almost all models have an error term, and that can be considered a latent variable.)

The first chapter of Loehlin's book 'Latent Variable Models: An Introduction to Factor, Path, and Structural Equation Analysis' talks about SEMs without latent variables.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.