I have a model with 4 binary exogenous , one endogenous/mediator (binary), and a latent endogenous (continuous). I didn't know Mplus needed info about categorical variables.
My original input: MODEL: F1 BY Y2@1 Y3 Y4*; F1 ON X3 X4 X5 Y1; Y1 ON X1 X2 X3 X4; MODEL INDIRECT: F1 IND Y1 X3; F1 IND Y1 X1; F1 IND Y1 X2; F1 IND Y1 X4; OUTPUT: SAMPSTAT STANDARDIZED RESIDUAL STANDARDIZED MODINDICES (3.84);
Now, I actually like and understand the results of this output. However, Y1 is a categorical. What is the best procedure?
My next attempt was: CATEGORICAL = Y1 X1 X3 X4; ANALYSIS: TYPE = GENERAL; ESTIMATOR=ML MODEL: F1 BY Y2@1 Y3 Y4*; F1 ON X1 X3 X4 Y1; Y1 ON X1 X2 X3 X4; OUTPUT: SAMPSTAT STANDARDIZED RESIDUAL;
The output from this is confusing. No indirect effects are allowed? I understand odds ratio, but the output isn't clear. Does the output below suggest that when X4 = 1, the odds that Y1 = 1 are 1.947. That is about twice as likely as when X4 = 0?
The CATEGORICAL option is for dependent variables only so you should remove the independent variables. When you ask for ML with categorical variables, you obtain logistic regression. If you remove ESTIMATOR=ML, you will obtain the default WLSMV estimator and probit regression.