Categorical indicators, WLS, and corr... PreviousNext
Mplus Discussion > Categorical Data Modeling >
Message/Author
 Anonymous posted on Wednesday, June 09, 2004 - 5:10 pm
hi,

i had been under the impression that, at least with older versions of mplus, it was not possible to allow for correlations between residual variances of categorical indicators in cfa models using the wls estimator. a previous discussion makes me wonder:

1. whether this is true at all and
2. whether this is true in version 3

thanks very much for any help.
 Linda K. Muthen posted on Wednesday, June 09, 2004 - 5:22 pm
With weighted least squares estimators, residual covariances have always been allowed in Mplus -- Version 1, Version 2, and Version 3.
 Anonymous posted on Thursday, August 11, 2005 - 7:26 pm
Hi, age is an independent variable (note: NOT dependent variable) in my model and I have transform it into 4 categorical variables:

(1)younger than 30; (2) 30 to 39; (3) 40 to 49; (4) older than 50.

When I run SEM with MPLUS, do I need to specify anything in the command, or do I need to omit one of them so that it can be taken as the comparison baseline? In STATA, it will automatically drop one of them and take it as the comparison baseline. I don't know how MPLUS deals with this issue.

Thanks!
 Linda K. Muthen posted on Friday, August 12, 2005 - 6:35 am
When you create dummy variables to represent a nominal or in your case ordinal categorical variable, you need one less dummy variable than the number of categories. Mplus does not automatically delete one dummy variable.
 Thomas Olino posted on Saturday, December 03, 2005 - 4:10 am
I just wanted to clarify one thing for presenting model diagrams appropriately.

Using WLSMS estimator and theta parameterization, for a model such as:

f1 BY u1-u4;
f2 BY u5-u8;
u1 WITH u5;
u2 WITH u6;
u3 WITH u7;
u4 WITH u8;

The covariance paths would be drawn from the residual error terms of the indicators, right?

Thanks.
 Linda K. Muthen posted on Saturday, December 03, 2005 - 6:07 am
Residual covariances would be specified for both the delta and theta parameterizations as curved double-sided arrows connecting two residuals.
 WJCAO posted on Tuesday, November 01, 2011 - 2:34 am
Hi, I wonder if you can tell me what's wrong with the following procedure.
I want to generate 4-liket data according to 2-factor CFA.
TITLE: categorical data
MONTECARLO: NAMES ARE y1-y10;
NOBSERVATIONS = 150;
NREPS = 500;
SEED = 53487;
generate = y1-y10(3);
categorical = y1-y10;
SAVE = cfa1.dat;
ANALYSIS:
ESTIMATOR = WLS;

Model population:
[y1$1*0 y2$1*0 y3$1*0 y4$1*0 y5$1*0 y6$1*0 y7$1*0 y8$1*0 y9$1*0 y10$1*0] ;
[y1$2*1 y2$2*1 y3$2*1 y4$2*1 y5$2*1 y6$2*1 y7$3*1 y8$2*1 y9$2*1 y10$2*1] ;
[y1$3*2 y2$3*2 y3$3*2 y4$3*2 y5$3*2 y6$3*2 y7$3*2 y8$3*2 y9$3*2 y10$3*2] ;
f1 BY y1-y5*.8;
f2 BY y6-y10*.8;
f1@1 f2@1;
y1-y10*.36;
f1 WITH f2*.25;
Model:
f1 BY y1-y5*.8;
f2 BY y6-y10*.8;
f1@1 f2@1;
y1-y10*.36;
f1 WITH f2*.25;

OUTPUT: TECH9;
 Linda K. Muthen posted on Tuesday, November 01, 2011 - 7:07 am
Please send the output and your license number to support@statmodel.com.
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: