Growth curve mediation
Message/Author
 Ashley D'Inverno posted on Friday, July 07, 2017 - 11:10 am
We are attempting to test the mediating effect of a slope outcome and slope mediator on an observed predictor. We used the TYPE = COMPLEX RANDOM code and got the following error messages:

THE ESTIMATED COVARIANCE MATRIX COULD NOT BE INVERTED.COMPUTATION COULD NOT BE COMPLETED IN ITERATION 116.CHANGE YOUR MODEL AND/OR STARTING VALUES.

THE MODEL ESTIMATION DID NOT TERMINATE NORMALLY DUE TO AN ERROR IN THE COMPUTATION. CHANGE YOUR MODEL AND/OR STARTING VALUES.

We used the RANDOM code because we are using age, not wave, as our time variable because our subjects are not the same age at each wave and the waves are not equally spaced. What suggestions do you have to overcome the error messages and get the model to converge?
 Bengt O. Muthen posted on Friday, July 07, 2017 - 4:33 pm
 Dennis Reidy posted on Friday, September 01, 2017 - 6:06 am
We have conducted a test of the mediating effect of a latent growth slope on the relationship between treatment dosage and the slope of the outcome. We used the Model Constraint to create a new Indirect parameter that is product of path A and path B using the following input syntax:

i_m i_y on X ;
s_y on X;
s_m on X (a1);
s_y on s_m (a2);

MODEL CONSTRAINT:
NEW(Indirect1);
Indirect1 = a1*a2;

Our question are

1) if the indirect parameter is significant does this mean that we have an indirect effect?

2) is the beta for this indirect parameter the amount of indirect effect?

3) if the regression parameter for the slope of y on x is still significant when including the indirect parameter is this partial mediation?

4) How do we determine what the amount of the direct effect is? Would this be the parameter of y on x minus the indirect parameter?
 Bengt O. Muthen posted on Friday, September 01, 2017 - 1:05 pm
You want to ask these general analysis questions on SEMNET.