Warning Message
Message/Author
 Kerrie Wilkins posted on Sunday, March 31, 2013 - 3:30 pm
Hi,
I am a beginner with Mplus and I am trying to run a CFA.

I have used this syntax:

VARIABLE:
NAMES ARE y1-y6;
MODEL:
f1 BY y1;
f1 BY y3;
f1 BY y5;
f2 BY y2;
f2 BY y4;
f2 BY y6;
OUTPUT:
SAMPSTAT; STAND; MOD; TECH4;

Then the modification indices suggested I free the Y2 WITH Y6 parameters.

When I do that I get a warning message:

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, OR A LINEAR DEPENDENCY AMONG MORE THAN TWO LATENT VARIABLES.CHECK THE TECH4 OUTPUT FOR MORE INFORMATION.PROBLEM INVOLVING VARIABLE F2.

In the tech4 section I have found that the correlations between the latent factors are 1.328.

The Chi Square is not significant. The CFI is .953, the TLI is .899, the RMSEA is .07 (CI - .000 to .139), and the SRMR is .04.

Should I disregard this modification index? I don't get this error message if I constrain Y2 WITH Y6. I don't know how to solve the problem. Could you help, please?

Kerrie
 Bengt O. Muthen posted on Monday, April 01, 2013 - 3:16 pm
If the residual correlation (Y2 WITH Y6) is significant and makes substantive sense, you have to ask why the factor correlation is inadmissible. It probably happens because your model is misspefied. For instance, do you have significant cross-loadings (an item loading on both factors)? You may want to explore this using EFA or BSEM.
 Kerrie Wilkins posted on Tuesday, April 02, 2013 - 8:55 pm
Thank you Dr. Muthen. I appreciate your response. I will examine it further using an EFA.