Corinna posted on Saturday, July 01, 2017 - 9:55 am
Dear Linda & Bengt, I have a question on LCGA: In contrast to GMM in LCGA we can only analyze the influence of the predictors on the class membership (= logistic regression). My question now is: If I conduct a conditional model (LCGA) and include the predictors directly in the model, do they then exert an influence on the growth curve themselves or is the inclusion of the predictors independent of the classes I extract and their characteristics? I tried to answer that question myself and calculated LCGAs with and without predictors with the same sample. The result was that the means of the growth factors did only differ slightly... but they differed. What does that mean for my question?
I would be very thankful for a short answer. With best regards
The class formation can change when adding predictors. Often this is due to direct effects from some covariates to some latent class indicators, that is, measurement non-in variance. If you want to avoid this change in class formation, you can take a 3-step approach. See our web notes 15 and 21 on our website.