DIFFTEST with EFA PreviousNext
Mplus Discussion > Exploratory Factor Analysis >
Message/Author
 John Dantzler posted on Tuesday, January 31, 2012 - 3:15 pm
I am attempting to run difftest to assess the difference between a 2-factor and 3-factor EFA with categorical data. I used the DIFFTEST option in the SAVEDATA command for the 3-factor solution and then ran a 2-factor solution with the DIFFTEST option in the ANALYSIS command. I am receiving no chi-square difference testing output, however. Any guidance would be appreciated.

VARIABLE:
NAMES ARE RES1-RES22;
USEVARIABLES ARE RES1-RES22;
CATEGORICAL ARE RES1-RES22;

ANALYSIS:
TYPE IS EFA 2 2;
ESTIMATOR IS WLSMV;
ITERATIONS = 1000;
CONVERGENCE = 0.00005;
ROTATION = GEO;
DIFFTEST = c:\fac3.dat;
 Linda K. Muthen posted on Tuesday, January 31, 2012 - 8:28 pm
The DIFFTEST option is not available with TYPE=EFA. You can do an EFA model using ESEM and use DIFFTEST there. See Example 5.24. ESEM without covariates and direct effects is EFA.
 John Dantzler posted on Tuesday, January 31, 2012 - 9:44 pm
Ah! Thank you.
 Felicia Gibson posted on Tuesday, March 05, 2013 - 8:02 pm
I am running a complex EFA with categorical and continuous data that is also nested (so I am using a cluster variable). I am using WLSMV.

I have two questions:

1. I am trying to determine which factor solution is best and normally would look at Chi-Square difference testing, but there is a note that says you cannot use the ch-square difference testing in the normal way and to use the DIFFTEST option. I tried doing this and it would not work? Is there another way to compare the factor solutions in an EFA?

2. I ran an EFA with option for MODINDICES and it gave me a Theta matrix? I have no idea what this means, or what I am supposed to do with that information. My goal was to use info from MODINDICES to make my model better before running a CFA?

Thank you!
 Linda K. Muthen posted on Wednesday, March 06, 2013 - 3:37 pm
1. Send the output and your license number to support@statmodel.com.

2. These are residual covariances. Ask for MODINDICES (ALL).
 Po-Yi Chen posted on Tuesday, December 10, 2013 - 2:47 pm
Dear Dr.Muthen,

you mentioned that we can use ESEM to do the chi square different test that based on WLSMV, I am wondering do you know or have published any article that specifically address the issue of conducing the chi corrected different test base on corrected chi squared in the context of ESEM?

thank you!
 Tihomir Asparouhov posted on Wednesday, December 11, 2013 - 2:37 pm
Po-Yi

The Chi-square test of fit in general does not account for parameter rotation because the the test of fit for an EFA or an ESEM model is based on the unrotated CFA solution. Thus the Chi-square test of fit is independent and unrelated to the rotation in any way, so there are no ESEM specifics for the chi-square test. The chi-sqaure test is essentially based on a CFA 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: