Anonymous posted on Thursday, December 09, 2004 - 8:21 am
cross-lagged panel (autoregressive simplex) models are often argued to be good at disentangling the causal order between two correlated processes. However, would it not be possible to address the same issue by specifying reciprocal effects between latent slopes in a simultaneous process model? Would there be major identification problesm with such an approach? What special concerns would need to be taken into account in terms of model specifciation to achieve unbiased estimates of the reciprocal paths?
Hi, I would like to model an autoregressive latent trajectory model as described by Curran/Bollen [Sociological Methods & Research, Vol. 32, No. 3, 336-383 (2004)]for a three wave panel with single indicators only for two constructs. Am I correct that it would possible with Mplus 3.0 to add non-linear constraints to the growth part of the model? Are there any examples available based on Mplus?
bmuthen posted on Saturday, February 26, 2005 - 11:51 pm
Yes, Mplus version 3 handles non-linear parameter constraints so this analysis should work fine. There are a couple of general examples in the version 3 User's Guide. And there is one example also on Mplus Discussion - although I can't put my finger on where right now.