gsem with fixed lclasses

[Hanna Hausmann]

I would like to have several gsem models estimated, each with additional control variables. Furthermore, I want to have two classes (a cluster analysis has identified two classes).

Unfortunately, I cannot find a way to keep the classes the same between the models.

Here is my code:

Model 1:
gsem (z_CCS_INFO_PRO <- z_PEB_new), lclass(C 2)

Model 2:
gsem (z_CCS_INFO_PRO <- z_PEB_new CCS_EVALUATION1 CCS_EVALUATION2 CCSK CCK heard), lclass(C 2)

How can I ensure that the same lclass is used in both models?

Additionally, is it possible to use my cluster (variable: clus1) who determine the two classes?

Thanks in Advance
Hanna

[Erik Ruzek]

Question: Why is it necessary to keep the individuals in the same latent class from model 1 to model 2?

The latent classes are not fixed in any sense. What the models are doing in very simple language is estimating a probability of an individual being in one of two classes. In model 2, you are giving the estimation routine more information (predictors) to make the classification. Based on those predictors, it is telling you that the probabilities of individuals being in one class or the other is shifting.

Perhaps you have done this already but if not, I would suggest reading Example 50g, 51g, and 52g of the sem manual very carefully.

[Hanna Hausmann]

Thanks for your reply!

We thought to control for the same classes would be a nice benefit (we also tried it without them) and to make it easier to compare coefficients between models.