The article examines tests for structural equation models with hidden variables and measurement errors. The commonly used chi square test has limitations in this context. New methods are proposed to address these challenges and improve the accuracy of model evaluation.