Multi-group CFA with ceiling effects ... PreviousNext
Mplus Discussion > Confirmatory Factor Analysis >
Message/Author
 Bethany Lee posted on Monday, October 09, 2006 - 9:08 am
We are conducting a 2-group CFA with a 13-item ordered categorical measure.

We are using WLSMV estimation since response options are positively skewed for at least one group (I had thought from the Flora & Curran, 2005 article that WLSMV was most appropriate for categorical data with floor or ceiling effects, although a post on this site suggested WLS or ML).

We have received the following error message:

*** ERROR
> Group 2 does not contain all values of categorical variable


It seems that the range for one group of respondents is different than the other group, but can the multi-group CFA still be conducted?

Do we need to change our estimation method?

Thanks for your input!
 Linda K. Muthen posted on Tuesday, October 10, 2006 - 8:22 am
You have two options. You can use maximum likelihood estimation and special features of the CATEGORICAL option which allow a different number of categories in each group. See the user's guide for details. Or you can collapse categories and use weighted least squares.
 Cheng-Hsien Li posted on Tuesday, April 10, 2012 - 12:49 pm
We are having the same issue that some groups do not contain all values of categorical variables.

Other than using ML with categorical option or collapsing categories with the use of WLS, is there any other option to deal with it?

Thanks.
 Linda K. Muthen posted on Tuesday, April 10, 2012 - 1:46 pm
Those are the only two options that I know of.
 Cheng-Hsien Li posted on Wednesday, April 11, 2012 - 12:53 pm
Thanks for your response.

We are trying to use ML with categorical option in a three-group CFA model, but it gives us an error message:

ALGORITHM = INTEGRATION is not available for multiple group analysis.
Try using the KNOWNCLASS option for TYPE = MIXTURE.

It seems that a multiple group CFA with ordered categorical outcomes can only be conducted by a mixture model with the KNOWNCLASS option while using ML estimation??
 Linda K. Muthen posted on Wednesday, April 11, 2012 - 1:12 pm
Yes, that is true.
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: