Introduction to Structural Equation Modeling with Latent Variables

Model Modifications

When you fit a model and the model fit is not satisfactory, you might want to know what you could do to improve the model. The LM (Lagrange multiplier) tests in PROC CALIS can help you improve the model fit by testing the potential free parameters in the model. To request the LM tests, you can use the MODIFICATION option in the PROC CALIS statement.

The LM test results contain lists of parameters, organized according to their types. In each list, the potential parameter with the greatest model improvement is shown first. Adding these new parameters improves the model fit approximately by the amount of the corresponding LM statistic.

Sometimes, researchers might have a target set of parameters they want to test in the LM tests. PROC CALIS offers a flexible way that you can customize the set the parameters for the LM tests. See the LMTESTS statement for details.

In addition, the Wald statistics produced by PROC CALIS suggest whether any parameters in your model can be dropped (or fixed to zero) without significantly affecting the model fit. You can request the Wald statistics with the MODIFICATION option in the PROC CALIS statement.