Model Fit with Complex Survey Data PreviousNext
Mplus Discussion > Multilevel Data/Complex Sample >
Message/Author
 Stefanie Hinkle posted on Monday, March 21, 2011 - 11:13 am
Hello. I am running a measured variable path analysis model with complex survey data. I am using replicate weights (jk2). I would like to obtain model fit information. Right now the program is only giving me the weighted root mean square residual. Is there any way to request the model chi square, AIC or RMSEA?

Below is the code that I am running.

Thank you very much for the help!

TITLE:
MEASURED VARIABLE PATH ANALYSIS
DATA:
FILE IS " ";

VARIABLE:
NAMES ARE V1- V9 re1-re6 p1- p3
ed1-ed4 w3r0 w3r1-w3r90;

USEVARIABLES
V1- V9 re2-re6 p2- p3 ed2-ed4
w3r0 w3r1-w3r90;
;


WEIGHT = w3r0;
REPWEIGHTS = w3r1-w3r90;
!SUBPOPULATION = V8 EQ 1;

MISSING ARE .;


CATEGORICAL ARE V7 ;


DATA:

LISTWISE = ON;

ANALYSIS:

TYPE= GENERAL COMPLEX ;

REPSE=JACKKNIFE2;

PARAMETERIZATION = THETA;

MODEL:

V3 ON V9 V4 V5 RE2-RE6 p2-p3 ed2-ed4 ;

V2 ON V3 V9 V4 V5 RE2-RE6 p2-p3 ed2-ed4 ;

V6 ON V2 V3 V9 V4 V5 RE2-RE6 p2-p3 ed2-ed4 ;

V7 ON V3 V6;

V1 ON V2 V3 V7 V6;

OUTPUT:

SAMPSTAT STANDARDIZED;
 Linda K. Muthen posted on Monday, March 21, 2011 - 1:02 pm
Fit statistics have not yet been developed for replicate weights.
 Stefanie Hinkle posted on Monday, March 21, 2011 - 1:04 pm
Ok. Thanks. That is very helpful to know!
 Stefanie Hinkle posted on Monday, March 21, 2011 - 1:40 pm
Hi Dr. Muthen,

If I use the stratification and cluster statements instead of the replication method is it possible to get the AIC?

Thanks,

Stefanie
 Linda K. Muthen posted on Monday, March 21, 2011 - 1:53 pm
Yes, then you will get all available fit statistics.
 slck.dgn@hotmail.com posted on Saturday, November 12, 2016 - 7:51 am
Hi Dr. Muthen,

after 5+ years, I would like to follow up: have we still not developed fit statistics for replicate weights?
 Tihomir Asparouhov posted on Monday, November 14, 2016 - 3:13 pm
You can use model constraints to test "univariate hypothesis" by forming new parameters such as the difference between two parameter. It is possible to do model test / wald test (for multivariate hypothesis) using standard methodology, but that is not implemented in Mplus yet.

I don't know of any new developments or publish articles on LRT with replicate weight. If you know of such developments please send the papers.
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: