R. Jackson posted on Friday, January 30, 2009 - 1:14 pm
Hello All, I am creating a path analysis with 5 observed variables, 2 dependent and 3 independent. In the model diagram the independent variables are shown to be related (double headed arrows). Do I need to model that relationship in my MODEL statement or does MPLUS assume the independent variables are related? If I do include that relationship in the MODEL statement, do I use the WITH command to relate them? Thanks.
No, you do not need to specify the relationships among observed exogenous variables in the MODEL command. The model is estimated conditioned on these variables. Their means, variances, and covariances are not model parameters. If you want these values, do a TYPE=BASIC; with no MODEL command.
I am attempting to run a path analysis with both parallel and serial mediators. With regard to the input file, do I use "ON" or "WITH" statements? The model fit seems to vary greatly depending on this aspect.
ON is for regression and WITH is for covariance. A mediator typically interacts with another predictor where the interaction is obtained using Define (for observed variables) or XWITH (for latent variables) to create the interaction. The interaction is then regressed on.
Thanks for the response, Dr. Muthen! Also, I am trying to include a coviate in this same model (path analysis with parallel and serial mediation). How do I enter the covariate? Does it need to be entered onto each variable? Then, would this covariate require an "ON" or "WITH" statement?