Simple constraints on multigroup mean...
Message/Author
 Anonymous posted on Tuesday, February 22, 2005 - 7:37 am
Hello Bengt and Linda,

I know the coding must be a simple extrapolate of some of the examples in the Mplus 3 User's Guide, but I am having difficulty finding the particular model contraints to use for the following model. I have a randomized clinical trial with 2 groups, measures over 4+ time points (depending on the DV). I am seeking to compare ONLY the latent mean of the slope for the two groups. I have not, however, seen the process for freeing everything else and constraining a latent mean in LGM. Any help would be much appreciated.

Regards,

Jason
 Linda K. Muthen posted on Saturday, February 26, 2005 - 4:31 pm
I assume that you are talking about the mean of the slope growth factor. I am assuming that your outcome is continuous. In a growth model, the mean of the slope growth factor is free across groups as the default. Following is how the model is specified so that the mean of the slope growth factor is constrained to be equal across groups.

MODEL:
i s | y1@0 y2@1 y3@2 y4@3;
[s] (1);
 Anonymous posted on Monday, February 28, 2005 - 8:22 am
Excellent - thanks!
 Friedrich Platz posted on Monday, October 01, 2018 - 11:15 am
Hello,
I conducted an experiment with four different conditions measuring reaction times. Now I would like to test two competing hypotheses:
H1a: learning_gr1 >(learning_gr2 = learning_gr3)>learning_control
H1blearning_gr1=learning_gr2=learning_gr3)>learning_control

To test which of both hypotheses would fit better on my data, I use a multiple group latent growth model.
My questions are:
(a) Is the following syntax correct for my hypotheses?
H1a:
MODEL:
i s | T1@0 T7@7 T14@14 T21@21 T28@28;
[i];
MODEL CONSTRAINT:
MODEL group.contr:
[i] (1);
[s] (2);
MODEL group.treat1:
[i] (3);
[s] (4);
MODEL group.treat2:
[i] (5);
[s] (6);
MODEL group.treat3:
[i] (5);
[s] (6);

H1b:
MODEL CONSTRAINT:
MODEL group.contr:
[i] (1);
[s] (2);

MODEL group.treat1:
[i] (3);
[s] (4);

MODEL group.treat2:
[i] (3);
[s] (4);

MODEL group.treat3:
[i] (3);
[s] (4);

 Friedrich Platz posted on Monday, October 01, 2018 - 11:16 am
Dear Bengt and Linda,
I had to split my questions (and excuse that smiley above)
:-)

(b) I think, I should use a model fit testing approach. How can I do this with MPLUS?
(c) Which test would you suggest for testing the hypothesis mentioned above to see wether the growth of one group differ (in the assumed direction) significantly from the other? And how can this be done in MPLUS?
(d) How can I free the slope parameter for the control group?

I would be very happy if you could give me some examples or advice!
Kind regards,
Friedrich
 Bengt O. Muthen posted on Tuesday, October 02, 2018 - 5:44 am
Your Model command can use the

Model G1:

type of approach for your G1-G4 groups. But you should not put Model Constraint within those statements - MC is not part of the Model command.

There is no > testing in Mplus, just testing of being the same.

 Friedrich Platz posted on Tuesday, October 02, 2018 - 5:53 am
Thank you for your reply. I will contact you via the support, thus it's not clear for me what the "Model G1: type of approach" is. Thanks for your offer, contacting you!
 Bengt O. Muthen posted on Tuesday, October 02, 2018 - 6:06 am
It's the approach you are (partly) showing:

MODEL group.contr:
[i] (1);
[s] (2);