Multi-group analyses PreviousNext
Mplus Discussion > Structural Equation Modeling >
Message/Author
 Marie Alice posted on Monday, November 24, 2014 - 10:55 pm
I am trying to run multi-group analyses, to see whether the path coefficients differ by groups. (I only have observed variables in the model.)
As step 1, I ran separate models for two groups. As step 2, I ran the two groups together with unequal path coefficients.It seemed to run okay until here.

However, as step 3, I tried to run two groups together setting the path coefficients to be equal to compare the Chi-square statistics with the previous model, and I got an error message.

variable:
names = ses score single urban_r income feduc meduc;
usevariables = single score income;
missing = all(9999);
grouping=urban_r(0=rural 1=urban);

analysis:
type = general;
estimator = ML;
model is nomeanstructure;
model:
score on income single(1-2);
income on single(3);
model indirect:
score ind income single(4);
model rural:
output:
stand(STDYX) mod(all);

error message:
*** WARNING in ANALYSIS command
MODEL=NOMEANSTRUCTURE is not allowed in conjunction with TYPE=MISSING.
Request for MODEL=NOMEANSTRUCTURE will be ignored.
*** ERROR in MODEL INDIRECT command
There must be 2 values given with the IND statement representing the X and X* values.
Problem with: SCORE IND SINGLE

I would greatly appreciate your help.
 Linda K. Muthen posted on Tuesday, November 25, 2014 - 10:16 am
You must include means in the model.

You cannot use labels in MODEL INDIRECT. (4) cannot be included.
 Marie Alice posted on Tuesday, November 25, 2014 - 10:33 pm
Linda,
Thank you very much for your response.
The model ran okay without the "model is nomeanstrucutre" and the (4) after the model indirect.

One additional question.
To compare the chi-square statistics across two models, for the model without equal constraints, I get df=0, value=0. But for the model with equal constraints, I get df=3,and value=17.174.

1. Why is df=0? my sample size is 14383, and I only have three parameters in the model as shown below.(The model should not be saturated.)

model:
score on income single;
income on single;
model indirect:
score ind income single;

2. Can I subtract these two chi-square statistics (even if one is zero) to see which model fits better?

Once again, thanks for bearing with my introductory questions.
 Linda K. Muthen posted on Wednesday, November 26, 2014 - 8:06 am
1. Please send the output and your license number to support@statmodel.com.

2. Yes.
Back to top
Add Your Message Here
Post:
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Password:
Options: Enable HTML code in message
Automatically activate URLs in message
Action: