SEM Workflow:
https://www.pymc.io/projects/examples/en/latest/case_studies/bayesian_sem_workflow.html:
Issue description
There is a minor issue in the interpretation of the B matrix specification. The orientation of the target -> predictors encoding is incorrect. The correct structure is inverted. The row indexes denote the target the column indexes the predictors
Proposed solution
Will add a fix to correct the orientation. The interpretation i this case doesn't change, it's how the mistake snuck through.