Discriminant Validity in Mplus PreviousNext
Mplus Discussion > Confirmatory Factor Analysis >
Message/Author
 yun-ci, ye posted on Tuesday, June 28, 2011 - 11:25 am
Dear Dr. Muthen:
I perform pairwise chi-squared difference test to exam difference between the unconstrained model(coach and gt is free) and constrained model(coach and gt is fixed to 1). My constrained model(coach and gt is fixed to 1) for Mplus command is as follows.

TITLE: CFA with Discriminant Validity
DATA: FILE IS c:/mplus/gsw.dat;
VARIABLE: NAMES ARE g1 g2 g3 g4 g5 c1 c2 c3 c4 c5 t1 t2 t3 t4 t5 h1 h2 h3;
ANALYSIS:
estimator = ml;
iterations = 5000;
MODEL:
gt by g1* g2 g3 g4 g5;
coach by c1* c2 c3 c4 c5;
ath by t1* t2 t3 t4 t5;
we by h1* h2 h3;

gt@1
coach@1
ath@1
we@1

coach with gt@1;

Output:
sampstat;
TECH4;

But I run the command, and show error message is :

ˇ§WARNING: THE LATENT VARIABLE COVARIANCE MATRIX (PSI) IS NOT POSITIVE DEFINITE. THIS COULD INDICATE A NEGATIVE VARIANCE/RESIDUAL VARIANCE FOR A LATENT VARIABLE, A CORRELATION GREATER OR EQUAL TO ONE BETWEEN TWO LATENT VARIABLES, ....PROBLEM INVOLVING VARIABLE COACHˇ¨.

I donˇ¦t know what happen is? Is the command wrong, or others?

I appreciate your response beforehand. Thanks.
 Linda K. Muthen posted on Wednesday, June 29, 2011 - 10:12 am
You should use MODEL TEST to test that the correlation is one.
 Luisa Walls posted on Sunday, May 27, 2012 - 2:02 am
Dr. Muthen

Could you tell me if the following syntax is correct to test the discriminant validity with the procedure mentioned above with the command MODEL TEST

f1 by y1* y2 y3;
f2 by y4* y5 y6;
f3 by y7* y8 y9;

f1 with f2 (a1); !covariance=1 in model test
f1 with f3;
f2 with f3;

f1@1; f2@1; f3@1;

Model test:
a1 =1;

Thank you.
 Linda K. Muthen posted on Monday, May 28, 2012 - 8:37 am
Testing that the correlation is not equal to one is one test of discriminant validity.
 Luisa Walls posted on Monday, May 28, 2012 - 11:27 am
Thank you, so a significative Wald test indicates that there is discriminant validity among these constructs.
 Linda K. Muthen posted on Monday, May 28, 2012 - 12:22 pm
Yes.
 Dodam Park posted on Wednesday, April 05, 2017 - 3:38 am
Dear Dr. Muthen
When testing discriminant validity of a scale I am interested in using another scale which doesn't satisfy the normality assumption, is there any serious problem? If it is, what kind of solution I can use in that situation?
 Bengt O. Muthen posted on Wednesday, April 05, 2017 - 3:41 pm
Ask SEMNET.
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: