Chi-square difference testing using MLMV PreviousNext
Mplus Discussion > Confirmatory Factor Analysis >
Message/Author
 mikec posted on Saturday, June 11, 2005 - 2:11 pm
I would like to do a chi-square difference test for MLMV estsimation. The instructions for doing so are clearly laid out in the Mplus manual. When I attempt to do so, however, I get the following error message:

*** WARNING in Savedata command DIFFTEST is valid only for estimator WLSMV. Request for DIFFTEST will be ignored.

Any ideas as to why?

Also, I apologize if this is a silly question, but I was wondering if it is possible to perform CFA analyses with single indicators using Mplus? I know you can do it in AMOS by plugging the reliability value of the measure into the error term of the indicator.

Thanks again for your excellent advice and suggestions.

Michael
 Linda K. Muthen posted on Monday, June 13, 2005 - 12:51 am
You may not be using the most recent version of Mplus, Verson 3.12. If you are and still have a problem, please send your output, data, and license number to support@statmodel.com.

You can use one indicator in Mplus. You would say

f BY y1@1;
y1@a:

where a is the residual variance and is a function of reliability.
 michaelc posted on Tuesday, June 14, 2005 - 11:23 pm
Dr. Muthen:

Thank you for your (as always) prompt and helpful reply. I hope you find some down time this summer!

I am using Mplus version 3.0. Just to be clear, even though the DIFFTEST option with MLMV is listed in the manual for 3.0, it is available only in version 3.12?

Is there an upgrade available from 3.0 to 3.12?

Thanks again.

Michael
 bmuthen posted on Wednesday, June 15, 2005 - 7:42 am
I think DIFFTEST was introduced with 3.0. And, yes, if you have 3.0 you can upgrade to 3.12 - see the web site.
 michaelc posted on Thursday, June 16, 2005 - 6:08 am
If DIFFTEST with MLMV is indeed available in Version 3.0 (as the manual states), I am unsure as to why I am getting the error message:

*** WARNING in Savedata command DIFFTEST is valid only for estimator WLSMV. Request for DIFFTEST will be ignored.

Is an upgrade to Version 3.1 necessary for this function? On your order page, I see options for upgrading from Version 2 to Version 3 but not between Version 3.0 and 3.12.

Thank you again for your assistance with this wonderful tool.

Michael
 Sanjoy posted on Thursday, June 16, 2005 - 6:03 pm
Michael ...go to this link http://www.statmodel.com/support/index.html .... say you have baught "version3.0, base program + mixture", click on the link (2nd from top) ...then you will see this link http://www.statmodel.com/support/mixture_update.html ...everything is written there (like you have to download an executibale file etc.)... oh! do NOT uninstall your existing program, i,e, version3
regards
 Anonymous posted on Tuesday, September 13, 2005 - 2:37 pm
I am running a CFA with 8 categorical variables, 1 factor.

here are the results:
TESTS OF MODEL FIT

Chi-Square Test of Model Fit

Value 120.295*
Degrees of Freedom 17**
P-Value 0.0000

* The chi-square value for MLM, MLMV, MLR, WLSM and WLSMV cannot be used for
chi-square difference tests.

CFI/TLI
CFI 0.968
TLI 0.983

Number of Free Parameters 8

RMSEA (Root Mean Square Error Of Approximation)

Estimate 0.048

SRMR (Standardized Root Mean Square Residual)
Value 0.039

WRMR (Weighted Root Mean Square Residual)
Value 1.519

How do I interpret these results i.e., make sense of these results since they do not give me a coherent "story" of model fit or misfit?

thanks.
 bmuthen posted on Tuesday, September 13, 2005 - 4:33 pm
That's a big topic that have been discussed in several places, including on SEMNET and lots of SEM lit. See also Yu's dissertation which is on our web site. A rough way to handle this is to place (much) less emphasis on chi-square and WRMR (WRMR is often acting differently than other fit indices we have found more recently). From this perspective your model would appear to fit rather well. But it is good to search for neighboring models that fit better.
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: