Anonymous posted on Tuesday, June 21, 2005 - 11:05 am
Fixed effects regression methods are sometimes used with longitudinal data to control for unmeasured and/or stable characteristics of individuals. In a HLM framework, this tends to be done by including a series of dummy coded variables to represent each subject (with one subject serving as the reference person) in the model. Can this approach or a similar approach be implemented in mplus?
When the standard fixed-effects and Mplus growth models were applied to analyze the same panel data, can I say that the latter is methodologically better because it enables to do more than what the former does? If so, exactly how SEM growth model incorporates the main strength of fixed-effects model (i.e., controlling for unobserved heterogeneity)? Or, should they be understood basically as alternative modeling approaches? If so, in what sense are they complementary to each other?