Path coefficients Invariance across g... PreviousNext
Mplus Discussion > Structural Equation Modeling >
Message/Author
 Jinxin ZHU posted on Thursday, August 29, 2013 - 10:19 am
I am now using Type=Imputation, to conduct multigroup multilevel path analysis based on plausible value. I encountered three problems.
1. When testing the equality of the path coefficient. E.g. Model test: F_1=M_1; F_2=M_2; I got a significant result. However, when I do the test separately in two run, each of them was not significant?
Which result should I rely on?

2. I tried DIFFTEST; however, I cannot find a proper method of estimation to run DIFFTEST and multigroup multilevel analysis in one run. Do you have any suggestion?

3.Is there any way to compute indirect effect when using "TYPE=Imputation" for plausible value?

Many thanks.
 Linda K. Muthen posted on Thursday, August 29, 2013 - 5:04 pm
1-2. MODEL TEST computes a joint test which is more powerful. You should use MODEL TEST.

3. Use the MODEL CONSTRAINT command.
 Jinxin ZHU posted on Saturday, October 18, 2014 - 12:35 pm
Dear Linda,

1-2. For Model Test, you mean LR test right?

I am now using PV (5 sets), and got that: "The loglikelihood cannot be used directly for chi-square testing with imputed data".
4. Would you please suggest how to use the loglikelihood based on PV for LR test?

Thank you very much.
 Jinxin ZHU posted on Saturday, October 18, 2014 - 12:47 pm
Dear Linda,
Sorry for the bothering.
Please ignore or delete the previous questions. I meant:

I want to compare nested models using PV (5 sets). So I should conducted a LR test, but "the loglikelihood cannot be used directly for chi-square testing with imputed data".

Would you please suggest how to use the loglikelihood based on PV for LR test?

Many thanks.
 Linda K. Muthen posted on Saturday, October 18, 2014 - 5:10 pm
See MODEL TEST in the user's guide. It is a Wald test. You can't use the loglikelihood for difference testing with imputed data.
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: