Simultaneaous MODEL TEST statements v... PreviousNext
Mplus Discussion > Growth Modeling of Longitudinal Data >
 Dayuma Vargas posted on Monday, December 03, 2012 - 6:30 pm
Hello there,

I have a latent growth curve model and wanted to see if the intercept and slope means differed between two groups, so I used the MODEL TEST command. When I run a test simultaneously looking at the two latent constructs, the Wald Test is significant:

0 = i1 - i2;
0 = s1 - s2;

However, when I run a MODEL TEST for each comparison separately, neither one is significant. I am not sure how to interpret this. Any advice?

Thank you.
 Tihomir Asparouhov posted on Monday, December 03, 2012 - 8:44 pm
You can interpret this as follows. The two models with 1 constraint fit the data well, however the more constraint model with 2 constraints no longer fits the data well.

You can also use LRT(likelihood ratio test) and BIC for inference.
 Dayuma Vargas posted on Tuesday, December 04, 2012 - 3:05 am
While I understand that interpretation, it does not really help me understand the pattern of differences I have found. Neither intercept nor slope differ between the two groups when tested independently BUT when tested together, there is a group difference identified (both by MODEL TEST and by looking at the chi-square difference test of nested models). Can I say that group 1 and group 2 differ in their model of growth? (The intercept and slope are significant. If the significant difference identified when testing the two latent means together, then group 1 shows a lower baseline value and a shallower growth over time than group 2).
 Bengt O. Muthen posted on Tuesday, December 04, 2012 - 5:09 pm
Testing both together tests whether the line is equal across groups and this is a stronger hypothesis than testing either intercept or slope. The joint test would be the same as the two separate ones only if the two tests were uncorrelated.
Back to top
Add Your Message Here
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Options: Enable HTML code in message
Automatically activate URLs in message