Open
Description
CFA and SEM Notebook:
https://www.pymc.io/projects/examples/en/latest/case_studies/CFA_SEM.html#bayesian-structural-equation-models:
Issue description
Here Gwangwoo Kim raised the issue that a pair sigma terms in this model are not enforced to be positive and this can throw an error when you try and fit the model using Advi methods.
https://discourse.pymc.io/t/log-likelihood-evaluation-of-pymc-model/16770/4
Expected output
Adding a constraint to the parameter to enforce positivity in the sigma terms the model still fits the data using mcmc as before, but the example is more robust.
Proposed solution
If applicable, explain possible solutions and workarounds.