Structural equational modeling is a widely used method ins reserach, mainly in the social sciences.
My reserach interest is in alternative approches to this field, aiming at
he following goals:
- Estimate models with non-linear relations
- Estimate models with non-normal latent and manifest variables
- Estimate values of latent vaiables
- Check assumptions of standard SEM such as indendency of error variables
- Allow implicative relations in models
Extending the traditional approach to polynomial models seems to be simpler than
published approaches.
- Polynomial SEM based on Isserlis' theorem. This
Preprint explains the technique and contains links to implementations.
|
A very general approach is possible by estimating the equations case by case.
|
|