Baseline model in CFI / TLI PreviousNext
Mplus Discussion > Categorical Data Modeling >
Message/Author
 Jaume Aguado posted on Thursday, June 17, 2004 - 5:04 am
Hi
The user's guide isn't clear for me as to which is the baseline model M-Plus uses when computing CFI/TLI (for ordinal data). Can you clarify it?

Thank you
 Linda K. Muthen posted on Thursday, June 17, 2004 - 7:05 am
The Mplus User's Guide does not contain technical information. As stated in the user's guide, the technical appendices can be found at www.statmodel.com. The baseline model is the model of free variances and means if appropriate.
 Jaume Aguado posted on Thursday, June 17, 2004 - 8:23 am
I've read the fit indices section of the technical appendices (that's what I was refering to) but to me it is not clear how the baseline model is defined.
The appendix just says "it has uncorrelated outcomes with unrestricted variances and unrestricted means or thresholds". But how is the rest of the model; factors and their relation, and the lambdas?

Perhaps this is a silly question, don't know.

Thanks
 Linda K. Muthen posted on Thursday, June 17, 2004 - 9:03 am
The baseline model doesn't have factor loadings etc. The only parameters that are estimated are the variances of the observed variables and if they are part of the model, the means or thresholds for categorical outcomes. That is all the baseline model is.
 Jaume Aguado posted on Thursday, June 17, 2004 - 9:07 am
Thank you again.
 Li Lin posted on Wednesday, April 07, 2010 - 12:33 pm
I am sorry for asking the same question after reading previous messages. What is the baseline when the dependent variables are categorical? For a measurement model with continuous dependent variable, to reproduce the baseline model I can specify all paths from latent variable to a dependent variable equal to 1 and variance of latent variable equal to 0. But when dependent variables are categorical, what specification I should give to thresholds to reproduce a baseline model? More specifically, for example below, could you give me necessary lines of codes for specifying baseline model?
Thanks,
Li

VARIABLE: NAMES ARE Gender SFDES01 SFDES02 SFDES03 SFDES04;
GROUPING IS gender (1=male 2=female);
USEVARIABLES ARE SFDES01 - SFDES04;
CATEGORICAL ARE SFDES01 - SFDES04;
MISSING are SFDES01 - SFDES04 (0);
MODEL: desire by SFDES01 - SFDES04;
 Linda K. Muthen posted on Wednesday, April 07, 2010 - 12:50 pm
Folloiwng are the baseline models used in Mplus:

Continuous dependent variables: means and variances for observed dependent variables; means, variances, and covariances for observed independent variables

Categorical dependent variables: thresholds for categorical observed dependent variables; means, variances, and covariances for observed independent variables

If your variables are binary, the model would be:

MODEL:
[SFDES01$1 - SFDES04$1];
 Li Lin posted on Monday, April 12, 2010 - 1:22 pm
Thanks, Linda. I understand the baseline model now, then what is the null model that baseline model is comparing to in the "Chi-Square Test of Model Fit for the Baseline Model"? How is the df for this test calculated?
 Bengt O. Muthen posted on Monday, April 12, 2010 - 1:32 pm
The baseline model is compared to the unrestricted H1 model.
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: