Model comparison PreviousNext
Mplus Discussion > Structural Equation Modeling >
Message/Author
 Christine Yu posted on Friday, May 09, 2014 - 7:54 am
Hi,

If I am doing a MGA SEM where the full model has an interaction term (used type = random) and the simple model has the interaction term (used type = general), how do I do a model comparison since the full model's output does not include Chi-square Test of Model Fit info?
 Christine Yu posted on Friday, May 09, 2014 - 8:12 am
Update questions:

Am I correct to say I have 3 df given the below output, and I can do a loglikelihood difference Chi Square test by subtracting the absolute value of the LL of my full model from the LL of my simple model and use the critical Chi-square of 3 df to determine significance or not?

As it stands, it looks to me that the interaction model is no worse a fitting model than the simple model?

Does this even make sense to do this "general" a model comparison for a MGA SEM where Roman Catholics are set up as the reference group to which Mainline Protestants and Evangelicals are compared?

Stine


MODEL FIT INFORMATION (simple model)

Number of Free Parameters 46

Loglikelihood

H0 Value -6486.941
H1 Value -6388.825


Information Criteria

Akaike (AIC) 13065.883
Bayesian (BIC) 13263.550
Sample-Size Adjusted BIC 13117.529
(n* = (n + 2) / 24)


MODEL FIT INFORMATION (Full model with interaction)

Number of Free Parameters 49

Loglikelihood

H0 Value -6484.429


Information Criteria

Akaike (AIC) 13066.858
Bayesian (BIC) 13277.416
Sample-Size Adjusted BIC 13121.872
(n* = (n + 2) / 24)
 Linda K. Muthen posted on Friday, May 09, 2014 - 8:56 am
-2 times the loglikelihood difference is distributed as chi-square.

Please limit posts to one window.
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: